_target_: diffusion_policy.workspace.train_diffusion_transformer_hybrid_workspace.TrainDiffusionTransformerHybridWorkspace checkpoint: save_last_ckpt: true save_last_snapshot: false topk: format_str: epoch={epoch:04d}-test_mean_score={test_mean_score:.3f}.ckpt k: 5 mode: max monitor_key: test_mean_score dataloader: batch_size: 64 num_workers: 8 persistent_workers: false pin_memory: true shuffle: true dataset_obs_steps: 2 ema: _target_: diffusion_policy.model.diffusion.ema_model.EMAModel inv_gamma: 1.0 max_value: 0.9999 min_value: 0.0 power: 0.75 update_after_step: 0 exp_name: default horizon: 10 keypoint_visible_rate: 1.0 logging: group: null id: null mode: online name: 2023.01.04-03.58.37_train_diffusion_transformer_hybrid_tool_hang_image_abs project: diffusion_policy_debug resume: true tags: - train_diffusion_transformer_hybrid - tool_hang_image_abs - default multi_run: run_dir: data/outputs/2023.01.04/03.58.37_train_diffusion_transformer_hybrid_tool_hang_image_abs wandb_name_base: 2023.01.04-03.58.37_train_diffusion_transformer_hybrid_tool_hang_image_abs n_action_steps: 8 n_latency_steps: 0 n_obs_steps: 2 name: train_diffusion_transformer_hybrid obs_as_cond: true optimizer: betas: - 0.9 - 0.95 learning_rate: 0.0001 obs_encoder_weight_decay: 1.0e-06 transformer_weight_decay: 0.001 past_action_visible: false policy: _target_: diffusion_policy.policy.diffusion_transformer_hybrid_image_policy.DiffusionTransformerHybridImagePolicy causal_attn: true crop_shape: - 216 - 216 eval_fixed_crop: true horizon: 10 n_action_steps: 8 n_cond_layers: 0 n_emb: 256 n_head: 4 n_layer: 8 n_obs_steps: 2 noise_scheduler: _target_: diffusers.schedulers.scheduling_ddpm.DDPMScheduler beta_end: 0.02 beta_schedule: squaredcos_cap_v2 beta_start: 0.0001 clip_sample: true num_train_timesteps: 100 prediction_type: epsilon variance_type: fixed_small num_inference_steps: 100 obs_as_cond: true obs_encoder_group_norm: true p_drop_attn: 0.3 p_drop_emb: 0.0 shape_meta: action: shape: - 10 obs: robot0_eef_pos: shape: - 3 robot0_eef_quat: shape: - 4 robot0_eye_in_hand_image: shape: - 3 - 240 - 240 type: rgb robot0_gripper_qpos: shape: - 2 sideview_image: shape: - 3 - 240 - 240 type: rgb time_as_cond: true shape_meta: action: shape: - 10 obs: robot0_eef_pos: shape: - 3 robot0_eef_quat: shape: - 4 robot0_eye_in_hand_image: shape: - 3 - 240 - 240 type: rgb robot0_gripper_qpos: shape: - 2 sideview_image: shape: - 3 - 240 - 240 type: rgb task: abs_action: true dataset: _target_: diffusion_policy.dataset.robomimic_replay_image_dataset.RobomimicReplayImageDataset abs_action: true dataset_path: data/robomimic/datasets/tool_hang/ph/image_abs.hdf5 horizon: 10 n_obs_steps: 2 pad_after: 7 pad_before: 1 rotation_rep: rotation_6d seed: 42 shape_meta: action: shape: - 10 obs: robot0_eef_pos: shape: - 3 robot0_eef_quat: shape: - 4 robot0_eye_in_hand_image: shape: - 3 - 240 - 240 type: rgb robot0_gripper_qpos: shape: - 2 sideview_image: shape: - 3 - 240 - 240 type: rgb use_cache: true val_ratio: 0.02 dataset_path: data/robomimic/datasets/tool_hang/ph/image_abs.hdf5 dataset_type: ph env_runner: _target_: diffusion_policy.env_runner.robomimic_image_runner.RobomimicImageRunner abs_action: true crf: 22 dataset_path: data/robomimic/datasets/tool_hang/ph/image_abs.hdf5 fps: 10 max_steps: 700 n_action_steps: 8 n_envs: 28 n_obs_steps: 2 n_test: 50 n_test_vis: 4 n_train: 6 n_train_vis: 2 past_action: false render_obs_key: sideview_image shape_meta: action: shape: - 10 obs: robot0_eef_pos: shape: - 3 robot0_eef_quat: shape: - 4 robot0_eye_in_hand_image: shape: - 3 - 240 - 240 type: rgb robot0_gripper_qpos: shape: - 2 sideview_image: shape: - 3 - 240 - 240 type: rgb test_start_seed: 100000 tqdm_interval_sec: 1.0 train_start_idx: 0 name: tool_hang_image_abs shape_meta: action: shape: - 10 obs: robot0_eef_pos: shape: - 3 robot0_eef_quat: shape: - 4 robot0_eye_in_hand_image: shape: - 3 - 240 - 240 type: rgb robot0_gripper_qpos: shape: - 2 sideview_image: shape: - 3 - 240 - 240 type: rgb task_name: tool_hang task_name: tool_hang_image_abs training: checkpoint_every: 50 debug: false device: cuda:0 gradient_accumulate_every: 1 lr_scheduler: cosine lr_warmup_steps: 1000 max_train_steps: null max_val_steps: null num_epochs: 3500 resume: true rollout_every: 50 sample_every: 5 seed: 42 tqdm_interval_sec: 1.0 use_ema: true val_every: 1 val_dataloader: batch_size: 64 num_workers: 8 persistent_workers: false pin_memory: true shuffle: false