| { |
| "type": "pi0_ecot", |
| "n_obs_steps": 1, |
| "normalization_mapping": { |
| "VISUAL": "IDENTITY", |
| "STATE": "IDENTITY", |
| "ACTION": "IDENTITY" |
| }, |
| "input_features": { |
| "observation.images.top": { |
| "type": "VISUAL", |
| "shape": [ |
| 3, |
| 224, |
| 224 |
| ] |
| }, |
| "observation.state": { |
| "type": "STATE", |
| "shape": [ |
| 7 |
| ] |
| } |
| }, |
| "output_features": { |
| "action": { |
| "type": "ACTION", |
| "shape": [ |
| 7 |
| ] |
| } |
| }, |
| "device": "cpu", |
| "use_amp": true, |
| "chunk_size": 4, |
| "n_action_steps": 4, |
| "max_state_dim": 32, |
| "max_action_dim": 32, |
| "resize_imgs_with_padding": [ |
| 224, |
| 224 |
| ], |
| "empty_cameras": 0, |
| "adapt_to_pi_aloha": false, |
| "use_delta_joint_actions_aloha": false, |
| "tokenizer_max_length": 72, |
| "proj_width": 1024, |
| "num_steps": 10, |
| "use_cache": true, |
| "attention_implementation": "eager", |
| "freeze_vision_encoder": true, |
| "train_expert_only": true, |
| "train_state_proj": true, |
| "paligemma_pretrained_path": null, |
| "optimizer_lr": 0.0003, |
| "optimizer_betas": [ |
| 0.9, |
| 0.999 |
| ], |
| "optimizer_eps": 1e-08, |
| "optimizer_weight_decay": 0.0, |
| "scheduler_warmup_steps": 500, |
| "scheduler_decay_steps": 30000, |
| "scheduler_decay_lr": 2.5e-05, |
| "tokenizer_pretrained": "google/paligemma-3b-pt-224", |
| "max_move_tokens": 8, |
| "num_z_tokens": 8, |
| "ecot_source": "future_action", |
| "ecot_embedding_mode": "online", |
| "qwen_offline_z_key": "observation.ecot.qwen_hidden", |
| "vjepa_offline_z_key": "observation.ecot.vjepa_hidden", |
| "ecot_vla_offline_z_key": "observation.ecot.openvla_hidden", |
| "offline_z_is_projected": false, |
| "offline_z_sidecar_path": null, |
| "offline_z_sidecar_index_key": "index", |
| "offline_z_sidecar_z_key": "z", |
| "z_projection_path": "runs/autoencoder/checkpoint.pt", |
| "gt_z_projection_path": null, |
| "vjepa_z_projection_path": "runs/vjepa_autoencoder/checkpoint.pt", |
| "ecot_vla_z_projection_path": "runs/ecot_vla_autoencoder/checkpoint.pt", |
| "gt_move_joiner": ". ", |
| "gt_move_prompt_prefix": "Move: ", |
| "qwen_base_model_id": "Qwen/Qwen3-VL-4B-Instruct", |
| "qwen_model_id": "StarVLA/Qwen3VL-OFT-Bridge-RT-1", |
| "qwen_checkpoint_file": "checkpoints/steps_5000_pytorch_model.pt", |
| "vjepa_model_id": "facebook/vjepa2-vitl-fpc64-256", |
| "vjepa_clip_key": "observation.images.image_0_future", |
| "vjepa_clip_frames": 16, |
| "ecot_vla_model_id": "Embodied-CoT/ecot-openvla-7b-bridge", |
| "ecot_vla_prompt_template": "Task: {instruction}\nReasoning:", |
| "ecot_vla_text_max_length": 256, |
| "online_ecot_llm_hidden_dim": 4096, |
| "future_action_z_projection_path": "runs/future_action_autoencoder_transformer/checkpoint.pt", |
| "future_action_horizon": 50, |
| "future_action_key": "future_action", |
| "future_action_pad_mask_key": "future_action_pad_mask", |
| "enable_cross_masking": true, |
| "cross_mask_prob_a_only": 0.05, |
| "cross_mask_prob_a_only_start": 0.0, |
| "cross_mask_prob_a_only_end": 0.05, |
| "cross_mask_prob_a_only_curriculum_steps": 15000, |
| "cross_mask_prob_z_only": 0.050000000000000044, |
| "cross_mask_prob_z_only_start": 0.95, |
| "cross_mask_prob_z_only_end": 0.05, |
| "cross_mask_prob_z_only_curriculum_steps": 15000, |
| "z_loss_weight": 1.0, |
| "z_noise_alpha": 2.0, |
| "train_full_denoise_every": 50, |
| "train_full_denoise_loss_weight": 0.1, |
| "train_full_denoise_grad_last_k": 1, |
| "train_full_denoise_on_z_only": false |
| } |
