{
    "dataset": {
        "repo_id": "cijerezg/yellow-car-offline-training-v2",
        "root": "outputs/jack_cube",
        "episodes": null,
        "max_episodes": null,
        "image_transforms": {
            "enable": false,
            "max_num_transforms": 3,
            "random_order": false,
            "tfs": {
                "brightness": {
                    "weight": 1.0,
                    "type": "ColorJitter",
                    "kwargs": {
                        "brightness": [
                            0.8,
                            1.2
                        ]
                    }
                },
                "contrast": {
                    "weight": 1.0,
                    "type": "ColorJitter",
                    "kwargs": {
                        "contrast": [
                            0.8,
                            1.2
                        ]
                    }
                },
                "saturation": {
                    "weight": 1.0,
                    "type": "ColorJitter",
                    "kwargs": {
                        "saturation": [
                            0.5,
                            1.5
                        ]
                    }
                },
                "hue": {
                    "weight": 1.0,
                    "type": "ColorJitter",
                    "kwargs": {
                        "hue": [
                            -0.05,
                            0.05
                        ]
                    }
                },
                "sharpness": {
                    "weight": 1.0,
                    "type": "SharpnessJitter",
                    "kwargs": {
                        "sharpness": [
                            0.5,
                            1.5
                        ]
                    }
                },
                "affine": {
                    "weight": 1.0,
                    "type": "RandomAffine",
                    "kwargs": {
                        "degrees": [
                            -5.0,
                            5.0
                        ],
                        "translate": [
                            0.05,
                            0.05
                        ]
                    }
                }
            }
        },
        "revision": null,
        "use_imagenet_stats": false,
        "video_backend": "pyav",
        "return_uint8": false,
        "streaming": false,
        "additional_offline_dataset_paths": []
    },
    "env": {
        "type": "gym_manipulator",
        "task": "Pick up the red truck and put it in the bowl",
        "fps": 30,
        "features": {
            "observation.images.side": {
                "type": "VISUAL",
                "shape": [
                    3,
                    128,
                    128
                ]
            },
            "observation.images.top": {
                "type": "VISUAL",
                "shape": [
                    3,
                    128,
                    128
                ]
            },
            "observation.state": {
                "type": "STATE",
                "shape": [
                    6
                ]
            },
            "action": {
                "type": "ACTION",
                "shape": [
                    6
                ]
            }
        },
        "features_map": {
            "observation.images.side": "observation.images.side",
            "observation.images.top": "observation.images.top",
            "observation.state": "observation.state",
            "action": "action"
        },
        "max_parallel_tasks": 1,
        "disable_env_checker": true,
        "robot": {
            "type": "so100_follower",
            "port": "/dev/ttyACM0",
            "disable_torque_on_disconnect": true,
            "max_relative_target": null,
            "cameras": {
                "side": {
                    "type": "opencv",
                    "fps": 30,
                    "width": 640,
                    "height": 480,
                    "index_or_path": 0,
                    "color_mode": "rgb",
                    "rotation": 0,
                    "warmup_s": 1,
                    "fourcc": null,
                    "backend": 0
                },
                "top": {
                    "type": "opencv",
                    "fps": 30,
                    "width": 640,
                    "height": 480,
                    "index_or_path": 2,
                    "color_mode": "rgb",
                    "rotation": 0,
                    "warmup_s": 1,
                    "fourcc": null,
                    "backend": 0
                }
            },
            "use_degrees": true,
            "id": "follower_arm_v2",
            "calibration_dir": null
        },
        "teleop": {
            "type": "so100_leader",
            "port": "/dev/ttyACM1",
            "use_degrees": true,
            "id": "leader_arm_v2",
            "calibration_dir": null
        },
        "processor": {
            "control_mode": "leader",
            "observation": {
                "add_joint_velocity_to_observation": false,
                "add_current_to_observation": false,
                "add_ee_pose_to_observation": false,
                "display_cameras": false
            },
            "image_preprocessing": {
                "crop_params_dict": null,
                "resize_size": [
                    224,
                    224
                ]
            },
            "gripper": {
                "use_gripper": true,
                "gripper_penalty": 0.0
            },
            "reset": {
                "fixed_reset_joint_positions": [
                    0.54,
                    -90.69,
                    99.55,
                    73.7,
                    -50.23,
                    42.71
                ],
                "reset_time_s": 10.0,
                "control_time_s": 200.0,
                "terminate_on_success": true
            },
            "inverse_kinematics": null,
            "reward_classifier": {
                "pretrained_path": null,
                "success_threshold": 0.5,
                "success_reward": 1.0
            },
            "max_gripper_pos": 30.0
        },
        "name": "real_robot"
    },
    "policy": {
        "type": "pi05_rl",
        "n_obs_steps": 1,
        "input_features": {
            "observation.images.side": {
                "type": "VISUAL",
                "shape": [
                    3,
                    224,
                    224
                ]
            },
            "observation.images.top": {
                "type": "VISUAL",
                "shape": [
                    3,
                    224,
                    224
                ]
            },
            "observation.state": {
                "type": "STATE",
                "shape": [
                    6
                ]
            }
        },
        "output_features": {
            "action": {
                "type": "ACTION",
                "shape": [
                    6
                ]
            }
        },
        "device": "cuda",
        "use_amp": false,
        "use_peft": false,
        "push_to_hub": true,
        "repo_id": "cijerezg/multi-task-toys-merged-v2",
        "private": null,
        "tags": null,
        "license": null,
        "pretrained_path": null,
        "paligemma_variant": "gemma_2b",
        "action_expert_variant": "gemma_300m",
        "dtype": "bfloat16",
        "chunk_size": 50,
        "n_action_steps": 50,
        "max_state_dim": 6,
        "max_action_dim": 32,
        "num_inference_steps": 5,
        "time_sampling_beta_alpha": 1.5,
        "time_sampling_beta_beta": 1.0,
        "time_sampling_scale": 0.999,
        "time_sampling_offset": 0.001,
        "min_period": 0.004,
        "max_period": 4.0,
        "rtc_config": {
            "enabled": true,
            "prefix_attention_schedule": "LINEAR",
            "max_guidance_weight": 10.0,
            "execution_horizon": 10,
            "debug": false,
            "debug_maxlen": 100
        },
        "image_resolution": [
            224,
            224
        ],
        "empty_cameras": 0,
        "use_dataset_stats": false,
        "normalization_mapping": {
            "VISUAL": "IDENTITY",
            "STATE": "MIN_MAX",
            "ENV": "MIN_MAX",
            "ACTION": "QUANTILES"
        },
        "action_tokenizer_name": "physical-intelligence/fast",
        "text_tokenizer_name": "google/paligemma-3b-pt-224",
        "max_action_tokens": 256,
        "fast_skip_tokens": 128,
        "max_decoding_steps": 200,
        "temperature": 0.0,
        "subtask_regeneration_interval": 1.5,
        "gradient_checkpointing": true,
        "compile_model": false,
        "compile_mode": "max-autotune",
        "freeze_vision_encoder": false,
        "train_expert_only": false,
        "knowledge_insulation": true,
        "action_encoding": "anchor",
        "loss_weight_flow": 1.0,
        "loss_weight_action_ce": 1.0,
        "loss_weight_subtask_ce": 1.0,
        "optimizer_lr": 2.5e-05,
        "optimizer_betas": [
            0.9,
            0.95
        ],
        "optimizer_eps": 1e-08,
        "optimizer_weight_decay": 0.1,
        "optimizer_grad_clip_norm": 1.0,
        "scheduler_warmup_steps": 1000,
        "scheduler_decay_steps": 30000,
        "scheduler_decay_lr": 2.5e-06,
        "tokenizer_max_length": 64,
        "task": "Pick up the orange cube and place it on the black X marker",
        "action_dim": 6,
        "drop_n_last_frames": 2,
        "critic_target_update_weight": 0.005,
        "num_critics": 1,
        "discount": 0.97,
        "reward_normalization_constant": 5.0,
        "terminal_failure_reward": -16.0,
        "online_steps": 20000,
        "online_buffer_capacity": 5000,
        "offline_buffer_capacity": 50000,
        "async_prefetch": false,
        "online_step_before_learning": 10,
        "policy_update_freq": 1,
        "grad_clip_norm": 2.0,
        "gradient_accumulation_steps": 16,
        "critic_lr": 5e-05,
        "actor_lr": 5e-05,
        "utd_ratio": 1,
        "actor_device": "cuda:0",
        "learner_device": "cuda:0",
        "use_separate_critic": true,
        "critic_llm_depth": 6,
        "critic_network_kwargs": {
            "hidden_dims": [
                256,
                256
            ],
            "activate_final": true
        },
        "trainable_params": {
            "vision_encoder_from_layer": {
                "vision_tower": 5,
                "multi_modal_projector": true
            },
            "language_from_layer": 0,
            "critic_language_from_layer": 5
        },
        "offline_steps": 10000,
        "inference_advantage": 1.0,
        "advantage_scaling": 0.2,
        "pi05_checkpoint": "outputs/pi05_base",
        "action_encoding_stats_path": "outputs/stats_jack/action_stats_anchor_jack_cube.pt",
        "dataset_stats": null,
        "storage_device": "cpu",
        "shared_encoder": false,
        "num_discrete_actions": null,
        "vision_encoder_name": null,
        "actor_learner_config": {
            "learner_host": "192.168.50.1",
            "learner_port": 50051,
            "policy_parameters_push_frequency": 180,
            "queue_get_timeout": 2
        },
        "concurrency": {
            "actor": "threads",
            "learner": "threads"
        }
    },
    "reward_model": null,
    "output_dir": "outputs/jack_pi05_full_offline_training_val_anchor_action_v1",
    "job_name": "default",
    "resume": false,
    "seed": 42,
    "cudnn_deterministic": false,
    "num_workers": 4,
    "batch_size": 8,
    "prefetch_factor": 4,
    "persistent_workers": true,
    "steps": 100000,
    "eval_freq": 20000,
    "log_freq": 20,
    "tolerance_s": 0.0001,
    "save_checkpoint": true,
    "save_freq": 100,
    "use_policy_training_preset": true,
    "optimizer": {
        "type": "multi_adam",
        "lr": 0.001,
        "weight_decay": 0.1,
        "grad_clip_norm": 10.0,
        "optimizer_groups": {
            "actor": {
                "lr": 5e-05
            },
            "critic": {
                "lr": 5e-05
            }
        }
    },
    "scheduler": null,
    "eval": {
        "n_episodes": 50,
        "batch_size": 22,
        "use_async_envs": true
    },
    "wandb": {
        "enable": true,
        "disable_artifact": true,
        "project": "so101_real_offline-v1",
        "entity": null,
        "notes": null,
        "run_id": "bfi8fh35",
        "mode": null,
        "offline_project": "so101_real_offline-v1",
        "add_tags": true
    },
    "peft": null,
    "sample_weighting": null,
    "rename_map": {},
    "checkpoint_path": null,
    "offline_output_dir": "outputs/jack_pi05_full_offline_training_val_anchor_action_v1",
    "offline_save_freq": 400,
    "buffer_cache_dir": null,
    "use_rerun": true,
    "video_logging_cameras": [
        "top",
        "wrist"
    ],
    "episode_logging_freq": 4,
    "episode_save_freq": 10,
    "probe_parameters": {
        "enable_actions": true,
        "enable_representations": true,
        "enable_attention": true,
        "enable_offline_inference": true,
        "enable_spatial_memorization": true,
        "enable_action_drift_jacobian": true,
        "enable_spatial_memorization_jacobian": true,
        "output_dir": "outputs/probe",
        "mode": "all",
        "max_episodes": 1,
        "n_frames_per_episode": 128,
        "offline_inference_n_frames": 5,
        "random_seed": 42,
        "timestep": 0.5,
        "ref_max_episodes": 2,
        "ref_n_frames_per_episode": 256,
        "action_pca_dims": 50,
        "repr_pca_dims": 100,
        "umap_n_neighbors": 15,
        "umap_min_dist": 0.1,
        "umap_seed": 42,
        "sites": "prefix,suffix",
        "ep_3d_a": 0,
        "ep_3d_b": 1,
        "subtask_injection": false,
        "validation_batch_size": 32,
        "attn_eval_episodes": null,
        "attn_eval_subsample": 2,
        "spatial_layers": "0,9,17",
        "spatial_n_frames": 32
    },
    "val_dataset_path": "outputs/annotated_dataset_validation",
    "val_split": 0.0,
    "val_freq": 400,
    "val_on_start": false,
    "skip_critic": true
}