| { |
| "type": "molmoact2", |
| "n_obs_steps": 1, |
| "input_features": { |
| "observation.images.image": { |
| "type": "VISUAL", |
| "shape": [ |
| 3, |
| 256, |
| 256 |
| ] |
| }, |
| "observation.images.wrist_image": { |
| "type": "VISUAL", |
| "shape": [ |
| 3, |
| 256, |
| 256 |
| ] |
| }, |
| "observation.state": { |
| "type": "STATE", |
| "shape": [ |
| 8 |
| ] |
| } |
| }, |
| "output_features": { |
| "action": { |
| "type": "ACTION", |
| "shape": [ |
| 8 |
| ] |
| } |
| }, |
| "device": "cuda", |
| "use_amp": false, |
| "use_peft": false, |
| "push_to_hub": false, |
| "repo_id": null, |
| "private": null, |
| "tags": null, |
| "license": null, |
| "pretrained_path": "allenai/MolmoAct2-LIBERO-LeRobot", |
| "checkpoint_path": "allenai/MolmoAct2-LIBERO", |
| "checkpoint_revision": null, |
| "checkpoint_force_download": false, |
| "trust_remote_code": true, |
| "chunk_size": 10, |
| "n_action_steps": 10, |
| "action_mode": "both", |
| "inference_action_mode": null, |
| "discrete_action_tokenizer": "allenai/MolmoAct2-FAST-Tokenizer", |
| "discrete_generation_max_steps": null, |
| "norm_tag": null, |
| "setup_type": "single franka robotic arm in libero", |
| "control_mode": "delta end-effector pose", |
| "image_keys": [ |
| "observation.images.image", |
| "observation.images.wrist_image" |
| ], |
| "normalize_language": true, |
| "add_setup_tokens": true, |
| "add_control_tokens": true, |
| "normalize_gripper": false, |
| "num_state_tokens": 256, |
| "max_sequence_length": null, |
| "expected_max_action_dim": 32, |
| "num_flow_timesteps": 8, |
| "flow_matching_cutoff": 1.0, |
| "flow_matching_time_offset": 0.001, |
| "flow_matching_time_scale": 0.999, |
| "flow_matching_beta_alpha": 1.0, |
| "flow_matching_beta_beta": 1.5, |
| "num_inference_steps": null, |
| "mask_action_dim_padding": true, |
| "enable_inference_cuda_graph": true, |
| "per_episode_seed": false, |
| "eval_seed": null, |
| "rtc_config": null, |
| "enable_lora_vlm": false, |
| "lora_rank": 64, |
| "lora_alpha": 16, |
| "lora_dropout": 0.05, |
| "lora_bias": "none", |
| "enable_lora_action_expert": false, |
| "enable_knowledge_insulation": false, |
| "freeze_embedding": true, |
| "train_action_expert_only": false, |
| "gradient_checkpointing": true, |
| "model_dtype": "bfloat16", |
| "softmax_auxiliary_loss": true, |
| "softmax_auxiliary_loss_scale": 0.0001, |
| "discrete_loss_token_weighting": "root_subsegments_root_tokens", |
| "optimizer_lr": 1e-05, |
| "optimizer_vit_lr": 5e-06, |
| "optimizer_connector_lr": 5e-06, |
| "optimizer_action_expert_lr": 5e-05, |
| "optimizer_betas": [ |
| 0.9, |
| 0.95 |
| ], |
| "optimizer_eps": 1e-06, |
| "optimizer_weight_decay": 0.0, |
| "optimizer_grad_clip_norm": 1.0, |
| "scheduler_warmup_steps": 200, |
| "scheduler_decay_steps": null, |
| "scheduler_decay_lr": 1e-06, |
| "normalization_mapping": { |
| "VISUAL": "IDENTITY", |
| "STATE": "QUANTILES", |
| "ACTION": "QUANTILES" |
| }, |
| "dataset_feature_names": { |
| "action": [ |
| "q_cmd_0", |
| "q_cmd_1", |
| "q_cmd_2", |
| "q_cmd_3", |
| "q_cmd_4", |
| "q_cmd_5", |
| "q_cmd_6", |
| "gripper_cmd" |
| ], |
| "observation.state": [ |
| "q_0", |
| "q_1", |
| "q_2", |
| "q_3", |
| "q_4", |
| "q_5", |
| "q_6", |
| "dq_0", |
| "dq_1", |
| "dq_2", |
| "dq_3", |
| "dq_4", |
| "dq_5", |
| "dq_6", |
| "gripper_width" |
| ] |
| } |
| } |
