| { |
| "vjepa_checkpoint": "/data/thinkwam/models/vjepa2/vjepa2_1_vitl_dist_vitG_384.pt", |
| "vjepa_dim": 1024, |
| "vjepa_image_size": 384, |
| "vjepa_patch_size": 16, |
| "vjepa_views": 3, |
| "qwen_path": "/data/thinkwam/models/Qwen2.5-0.5B", |
| "qwen_dim": 896, |
| "qwen_max_length": 512, |
| "qwen_action_tokens": 64, |
| "qwen_lora": true, |
| "qwen_lora_rank": 32, |
| "qwen_lora_alpha": 64, |
| "qwen_lora_dropout": 0.1, |
| "qwen_lora_target_modules": "all-linear", |
| "qwen_attn_implementation": "flash_attention_2", |
| "qwen_visual_tokens": 1088, |
| "qwen_hidden_only": true, |
| "action_context_dim": 1024, |
| "conditioning_mlp_hidden": 2048, |
| "online_vjepa": true, |
| "compile_vjepa": true, |
| "vjepa_compile_mode": "default", |
| "compile_action_dit": true, |
| "action_dit_compile_mode": "reduce-overhead", |
| "text_max_length": 128, |
| "visual_token_count": 1152, |
| "jepa_horizons": [ |
| 8, |
| 16, |
| 32 |
| ], |
| "jepa_token_counts": [ |
| 288, |
| 144, |
| 72 |
| ], |
| "jepa_context_dim": 1024, |
| "jepa_observation_tokens": 1152, |
| "jepa_layers": 2, |
| "jepa_heads": 8, |
| "jepa_loss_weight": 0.5, |
| "sigreg_loss_weight": 0.0, |
| "sigreg_num_projections": 128, |
| "sigreg_num_knots": 17, |
| "sigreg_t_max": 3.0, |
| "memory_dim": 1024, |
| "memory_slots": 16, |
| "memory_observation_tokens": 256, |
| "memory_action_history": 16, |
| "memory_layers": 2, |
| "memory_heads": 8, |
| "action_dim": 14, |
| "proprio_dim": 14, |
| "action_horizon": 32, |
| "action_dit_hidden": 1024, |
| "action_dit_layers": 16, |
| "action_dit_heads": 32, |
| "action_dit_inner_dim": 1536, |
| "action_dit_future_tokens": 32, |
| "flow_steps": 4, |
| "action_loss_weight": 1.0, |
| "batch_size": 20, |
| "gradient_accumulation_steps": 1, |
| "learning_rate": 1e-05, |
| "min_learning_rate": 2e-06, |
| "weight_decay": 0.0, |
| "warmup_ratio": 0.02, |
| "max_steps": 25000, |
| "num_workers": 8, |
| "dataloader_prefetch_factor": 2, |
| "dataloader_persistent_workers": true, |
| "dataloader_pin_memory": true, |
| "dataloader_worker_threads": 1, |
| "mixed_precision": "bf16", |
| "gradient_checkpointing": false, |
| "checkpoint_save_limit": 5, |
| "optimizer_beta1": 0.9, |
| "optimizer_beta2": 0.95, |
| "eval_every": 5000, |
| "eval_batches": 8, |
| "eval_fraction": 0.1, |
| "tensorboard": true, |
| "tensorboard_histogram_every": 100, |
| "open_loop_eval": true, |
| "save_best_checkpoint": true, |
| "closed_loop_after": false, |
| "closed_loop_suite": "all", |
| "closed_loop_world_size": 8, |
| "closed_loop_tasks": -1, |
| "closed_loop_episodes": 50, |
| "closed_loop_replan_steps": 20, |
| "closed_loop_max_steps": 520, |
| "closed_loop_save_videos": false, |
| "seed": 7, |
| "dataset": "robotwin2.0", |
| "dataset_root": "/data/thinkwam/datasets/robotwin2.0-lerobot-v3.0/clean", |
| "feature_cache": "", |
| "output_dir": "/data/thinkwam/runs/robotwin2_150k_short_25k_20261101", |
| "robotwin_frame_cache": "/data/thinkwam/datasets/thinkwam/robotwin2_clean_rgb384", |
| "robotwin_manifest": "manifest_3view.json", |
| "robotwin_stats_path": "/data/thinkwam/datasets/robotwin2.0-lerobot-v3.0/clean/dataset_stats.json", |
| "affordance_enabled": false, |
| "phase_enabled": false, |
| "vjepa_trainable": false, |
| "vjepa_adapter_enabled": true, |
| "vjepa_adapter_dim": 256, |
| "vjepa_adapter_learning_rate": 3e-06, |
| "affordance_decoder_dim": 256, |
| "affordance_decoder_layers": 2, |
| "affordance_decoder_heads": 8, |
| "affordance_channels": 1, |
| "affordance_loss_weight": 0.0, |
| "affordance_dice_weight": 0.5, |
| "affordance_loss_ramp_steps": 3000, |
| "affordance_target_key": "affordance", |
| "require_affordance_targets": false, |
| "video_model_dim": 1024, |
| "video_model_layers": 2, |
| "video_model_heads": 8, |
| "video_context_tokens": 256, |
| "video_residual_loss_weight": 0.8, |
| "action_refinement_loss_weight": 1.4, |
| "first_action_loss_weight": 1.2, |
| "video_condition_steps": 4, |
| "video_condition_train_steps": 1, |
| "action_condition_scale": 1.0, |
| "p1_stage": "joint", |
| "p0_checkpoint": "", |
| "freeze_qwen_steps": 0, |
| "freeze_vjepa_steps": 0, |
| "freeze_memory_steps": 0, |
| "affordance_memory_start_step": 5000, |
| "affordance_memory_ramp_steps": 5000, |
| "affordance_memory_max_scale": 0.05, |
| "affordance_confidence_threshold": 0.7, |
| "affordance_confidence_temperature": 0.1, |
| "affordance_upstream_start_step": 5000, |
| "affordance_upstream_ramp_steps": 5000, |
| "affordance_upstream_max_scale": 0.05, |
| "memory_learning_rate": 5e-06, |
| "qwen_lora_learning_rate": 2e-06, |
| "phase_num_classes": 5, |
| "phase_target_key": "phase", |
| "require_phase_targets": false, |
| "phase_loss_weight": 0.0, |
| "phase_loss_ramp_steps": 3000, |
| "phase_memory_start_step": 5000, |
| "phase_memory_ramp_steps": 5000, |
| "phase_memory_max_scale": 0.05, |
| "phase_upstream_start_step": 5000, |
| "phase_upstream_ramp_steps": 5000, |
| "phase_upstream_max_scale": 0.1, |
| "plus_dataset_root": "", |
| "plus_mixture_fraction": 0.25, |
| "plus_augmentations": false, |
| "plus_augmentation_probability": 0.0, |
| "plus_illumination_probability": 0.5, |
| "plus_texture_probability": 0.35, |
| "plus_camera_probability": 0.25, |
| "plus_noise_probability": 0.25, |
| "plus_object_position_probability": 0.25, |
| "plus_max_translation": 12, |
| "plus_noise_std": 0.025, |
| "plus_category_manifest": "", |
| "plus_language_category_factor": 1.0, |
| "plus_robot_initial_state_factor": 1.0, |
| "plus_initial_state_quantile": 0.7, |
| "task_sampling_enabled": true, |
| "task_sampling_goal_long_factor": 2, |
| "task_sampling_hard_factor": 2, |
| "target_manifest": "/data/thinkwam/datasets/thinkwam/robotwin2_150k_short_50c20r/manifest.json", |
| "random_dataset_root": "/data/thinkwam/datasets/robotwin2.0-lerobot-v3.0/random", |
| "base_checkpoint": "/data/thinkwam/runs/robotwin2_focused_resume_90k_to150k/checkpoints/checkpoint_step_0150000.pt", |
| "expected_global_batch": 80, |
| "base_step": 150000, |
| "transition_step": 20000, |
| "warmup_steps": 500, |
| "lr_floor_ratio": 0.2, |
| "checkpoint_every": 5000, |
| "clean_mmap_episodes": 4, |
| "label_lru_episodes": 32, |
| "decode_timeout": 30.0, |
| "robotwin_tasks": [ |
| "adjust_bottle", |
| "beat_block_hammer", |
| "blocks_ranking_rgb", |
| "blocks_ranking_size", |
| "click_alarmclock", |
| "click_bell", |
| "dump_bin_bigbin", |
| "grab_roller", |
| "handover_block", |
| "handover_mic", |
| "hanging_mug", |
| "lift_pot", |
| "move_can_pot", |
| "move_pillbottle_pad", |
| "move_playingcard_away", |
| "move_stapler_pad", |
| "open_laptop", |
| "open_microwave", |
| "pick_diverse_bottles", |
| "pick_dual_bottles", |
| "place_a2b_left", |
| "place_a2b_right", |
| "place_bread_basket", |
| "place_bread_skillet", |
| "place_burger_fries", |
| "place_can_basket", |
| "place_cans_plasticbox", |
| "place_container_plate", |
| "place_dual_shoes", |
| "place_empty_cup", |
| "place_fan", |
| "place_mouse_pad", |
| "place_object_basket", |
| "place_object_scale", |
| "place_object_stand", |
| "place_phone_stand", |
| "place_shoe", |
| "press_stapler", |
| "put_bottles_dustbin", |
| "put_object_cabinet", |
| "rotate_qrcode", |
| "scan_object", |
| "shake_bottle", |
| "shake_bottle_horizontally", |
| "stack_blocks_three", |
| "stack_blocks_two", |
| "stack_bowls_three", |
| "stack_bowls_two", |
| "stamp_seal", |
| "turn_switch" |
| ] |
| } |
|
|
