| { |
| "checkpoint": "", |
| "device": "cuda:0", |
| "init_mode": "scratch", |
| "base_checkpoint": null, |
| "state_dim": 14, |
| "action_dim": 14, |
| "n_cameras": 3, |
| "norm_stats_checkpoint": null, |
| "n_embodiments": 1, |
| "max_state_dim": 32, |
| "max_action_dim": 32, |
| "prefix_format": "raw", |
| "model": { |
| "type": "bridgequant_async_vla" |
| }, |
| "backbone": { |
| "type": "tiny_vl", |
| "model_name": "prajjwal1/bert-mini", |
| "num_layers": 4, |
| "train_mode": "finetune", |
| "language_first": false, |
| "kwargs": { |
| "load_weights": true, |
| "vision_model": "facebook/dinov3-vits16-pretrain-lvd1689m", |
| "hidden_size": 512, |
| "image_size": 256, |
| "n_cameras": 3, |
| "num_heads": 8, |
| "mlp_ratio": 4.0, |
| "mask_mode": "m1", |
| "freeze_text": true |
| } |
| }, |
| "expert": { |
| "type": "bridgequantasyncvla_expert", |
| "hidden_size": 512, |
| "depth": 6, |
| "num_heads": 8, |
| "mlp_ratio": 4.0, |
| "moh": { |
| "enabled": false, |
| "horizons": [], |
| "stride": 0, |
| "gate_temperature": 1.0, |
| "ind_weight": 1.0, |
| "bal_weight": 0.001, |
| "ind_reduction": "sum" |
| }, |
| "kwargs": { |
| "chunk_index_features": true |
| } |
| }, |
| "action_head": { |
| "type": "bridge_flow_matching", |
| "action_horizon": 16, |
| "n_action_steps": 8, |
| "num_steps": 1, |
| "gripper_bottleneck": false, |
| "bottleneck_residual_lambda": 0.01, |
| "bottleneck_residual_init_alpha": 0.01, |
| "kwargs": {} |
| }, |
| "proprio": { |
| "proprio_to_vlm": false, |
| "proprio_to_policy": true, |
| "state_dropout": 0.0, |
| "cycle_count_in_state": false, |
| "kwargs": {} |
| }, |
| "temporal": { |
| "n_obs_steps": 1, |
| "enabled": false, |
| "type": "retnet", |
| "window_size": 1, |
| "context_refresh_every": 8, |
| "context_validity": true, |
| "memory_age_aug": 0.0, |
| "retention_chunk_size": 0, |
| "decay_min": 0.03, |
| "decay_max": 1.0, |
| "burnin_steps": 0, |
| "history_noise_std": 0.0, |
| "history_noise_rho": 0.0, |
| "history_source": "executed_action", |
| "context_refresh_choices": null, |
| "drift_loss_weight": 0.0, |
| "reinjection_mode": "additive", |
| "refresh_proprio_delta_threshold": null, |
| "refresh_action_magnitude_threshold": null, |
| "refresh_frame_diff_threshold": null, |
| "refresh_instruction_change": true, |
| "d_model": 256, |
| "num_heads": 4, |
| "control_dt": 0.0667, |
| "kwargs": {}, |
| "stream_units": 1, |
| "stream_stride": 5, |
| "stream_stride_jitter": 2 |
| }, |
| "bridge": { |
| "enabled": true, |
| "num_queries": 96, |
| "dim": 512, |
| "tap_layers": [ |
| 0, |
| 1, |
| 2, |
| 3 |
| ], |
| "rank": 256, |
| "min_grounding_confidence": null, |
| "gate_init": 0.3, |
| "z_bridge_pos_embed": true, |
| "zero_z_bridge": false, |
| "kwargs": {} |
| }, |
| "roles": { |
| "entity": 24, |
| "goal": 24, |
| "embodiment": 24, |
| "progress": 12, |
| "global_context": 12, |
| "alignment": 0, |
| "alignment_asymmetric_injection": false, |
| "alignment_asymmetric_scale": 1.0, |
| "plan_film": false, |
| "plan_film_scale": 1.0, |
| "plan_film_state": false, |
| "lang_kv_to_expert": false, |
| "mem_kv_to_expert": false, |
| "masked_groups": [], |
| "routing": null, |
| "routing_rows": null, |
| "spatial_binding_weight": 0.0, |
| "spatial_binding_hidden": 256, |
| "spatial_binding_huber_delta": null, |
| "alignment_offset_weight": 0.0, |
| "alignment_source_group": null, |
| "alignment_margin_weight": 0.0, |
| "alignment_margin_epsilon_cm": 2.5, |
| "alignment_huber_delta": 0.1, |
| "alignment_hidden": 256, |
| "alignment_phase_beta": 2.0, |
| "alignment_phase_sigma": 2.0, |
| "alignment_obj_window_frames": null, |
| "alignment_stop_gradient": false, |
| "competitive_read_groups": [], |
| "attn_diversity_weight": 0.0, |
| "attn_entropy_weight": 0.0, |
| "attn_shaping_groups": [] |
| }, |
| "aux_losses": [ |
| { |
| "type": "smoothness", |
| "weight": 0.05, |
| "kwargs": {} |
| } |
| ], |
| "loss": { |
| "zero_padded_noise": true, |
| "mask_padded_dims": true, |
| "first_action_weight": 1.0, |
| "direction_commit_weight": 0.0, |
| "direction_commit_tau": 0.15, |
| "direction_commit_mode": "hard", |
| "gripper_bce_weight": 0.0, |
| "gripper_bce_logit_scale": 4.0, |
| "task_balance_alpha": 0.25, |
| "task_balance_mode": "excess", |
| "task_balance_ema": 0.99, |
| "task_balance_clip": 4.0, |
| "phase_transition_weight": 3.0, |
| "phase_transition_radius": 2, |
| "aux_exclude_contact_window": false |
| }, |
| "anchor": { |
| "enabled": false, |
| "layers": [ |
| 7, |
| 8, |
| 9, |
| 10, |
| 11 |
| ], |
| "span": "lang", |
| "weight": 0.05 |
| }, |
| "embodiment": { |
| "name": "aloha_agilex", |
| "embodiment_id": 0, |
| "action_names": [ |
| "left_joint_0", |
| "left_joint_1", |
| "left_joint_2", |
| "left_joint_3", |
| "left_joint_4", |
| "left_joint_5", |
| "left_gripper", |
| "right_joint_0", |
| "right_joint_1", |
| "right_joint_2", |
| "right_joint_3", |
| "right_joint_4", |
| "right_joint_5", |
| "right_gripper" |
| ], |
| "xyz_dims": [], |
| "rot_dims": [], |
| "gripper_dim": null, |
| "relative_action_dims": [ |
| 0, |
| 1, |
| 2, |
| 3, |
| 4, |
| 5, |
| 7, |
| 8, |
| 9, |
| 10, |
| 11, |
| 12 |
| ], |
| "gripper_policy_open_value": 1.0, |
| "gripper_policy_close_value": 0.0, |
| "gripper_env_open_value": -1.0, |
| "gripper_env_close_value": 1.0 |
| }, |
| "model_type": "bridgequant_async_vla", |
| "bridgequant_vla_format_version": 1, |
| "step": 80000 |
| } |
