DorianAtSchool/smolvla_autoscript_executor_base_scripted

1{2    "dataset": {3        "repo_id": "DorianAtSchool/color_relay_executor_single_scripted",4        "root": null,5        "episodes": null,6        "image_transforms": {7            "enable": false,8            "max_num_transforms": 3,9            "random_order": false,10            "tfs": {11                "brightness": {12                    "weight": 1.0,13                    "type": "ColorJitter",14                    "kwargs": {15                        "brightness": [16                            0.8,17                            1.218                        ]19                    }20                },21                "contrast": {22                    "weight": 1.0,23                    "type": "ColorJitter",24                    "kwargs": {25                        "contrast": [26                            0.8,27                            1.228                        ]29                    }30                },31                "saturation": {32                    "weight": 1.0,33                    "type": "ColorJitter",34                    "kwargs": {35                        "saturation": [36                            0.5,37                            1.538                        ]39                    }40                },41                "hue": {42                    "weight": 1.0,43                    "type": "ColorJitter",44                    "kwargs": {45                        "hue": [46                            -0.05,47                            0.0548                        ]49                    }50                },51                "sharpness": {52                    "weight": 1.0,53                    "type": "SharpnessJitter",54                    "kwargs": {55                        "sharpness": [56                            0.5,57                            1.558                        ]59                    }60                },61                "affine": {62                    "weight": 1.0,63                    "type": "RandomAffine",64                    "kwargs": {65                        "degrees": [66                            -5.0,67                            5.068                        ],69                        "translate": [70                            0.05,71                            0.0572                        ]73                    }74                }75            }76        },77        "revision": null,78        "use_imagenet_stats": true,79        "video_backend": "pyav",80        "streaming": false81    },82    "env": null,83    "policy": {84        "type": "smolvla",85        "n_obs_steps": 1,86        "input_features": {87            "observation.state": {88                "type": "STATE",89                "shape": [90                    691                ]92            },93            "observation.images.camera1": {94                "type": "VISUAL",95                "shape": [96                    3,97                    256,98                    25699                ]100            },101            "observation.images.camera2": {102                "type": "VISUAL",103                "shape": [104                    3,105                    256,106                    256107                ]108            },109            "observation.images.camera3": {110                "type": "VISUAL",111                "shape": [112                    3,113                    256,114                    256115                ]116            }117        },118        "output_features": {119            "action": {120                "type": "ACTION",121                "shape": [122                    7123                ]124            }125        },126        "device": "cuda",127        "use_amp": false,128        "use_peft": false,129        "push_to_hub": true,130        "repo_id": "DorianAtSchool/smolvla_autoscript_executor_base_scripted",131        "private": null,132        "tags": null,133        "license": null,134        "pretrained_path": "DorianAtSchool/smolvla-combined-finetune-automated-scripted",135        "chunk_size": 50,136        "n_action_steps": 50,137        "normalization_mapping": {138            "VISUAL": "IDENTITY",139            "STATE": "MEAN_STD",140            "ACTION": "MEAN_STD"141        },142        "max_state_dim": 32,143        "max_action_dim": 32,144        "resize_imgs_with_padding": [145            512,146            512147        ],148        "empty_cameras": 0,149        "adapt_to_pi_aloha": false,150        "use_delta_joint_actions_aloha": false,151        "tokenizer_max_length": 48,152        "num_steps": 10,153        "use_cache": true,154        "freeze_vision_encoder": true,155        "train_expert_only": false,156        "train_state_proj": true,157        "optimizer_lr": 0.0001,158        "optimizer_betas": [159            0.9,160            0.95161        ],162        "optimizer_eps": 1e-08,163        "optimizer_weight_decay": 1e-10,164        "optimizer_grad_clip_norm": 10.0,165        "scheduler_warmup_steps": 1000,166        "scheduler_decay_steps": 30000,167        "scheduler_decay_lr": 2.5e-06,168        "vlm_model_name": "HuggingFaceTB/SmolVLM2-500M-Video-Instruct",169        "load_vlm_weights": true,170        "add_image_special_tokens": false,171        "attention_mode": "cross_attn",172        "prefix_length": 0,173        "pad_language_to": "max_length",174        "num_expert_layers": 0,175        "num_vlm_layers": 16,176        "self_attn_every_n_layers": 2,177        "expert_width_multiplier": 0.75,178        "min_period": 0.004,179        "max_period": 4.0,180        "rtc_config": null181    },182    "output_dir": "outputs/finetune/smolvla_autoscript_executor_base_scripted",183    "job_name": "smolvla_autoscript_executor_base_scripted",184    "resume": false,185    "seed": 1000,186    "num_workers": 4,187    "batch_size": 32,188    "steps": 50000,189    "eval_freq": 20000,190    "log_freq": 200,191    "tolerance_s": 0.0001,192    "save_checkpoint": true,193    "save_freq": 20000,194    "use_policy_training_preset": true,195    "optimizer": {196        "type": "adamw",197        "lr": 0.0001,198        "weight_decay": 1e-10,199        "grad_clip_norm": 10.0,200        "betas": [201            0.9,202            0.95203        ],204        "eps": 1e-08205    },206    "scheduler": {207        "type": "cosine_decay_with_warmup",208        "num_warmup_steps": 1000,209        "num_decay_steps": 30000,210        "peak_lr": 0.0001,211        "decay_lr": 2.5e-06212    },213    "eval": {214        "n_episodes": 50,215        "batch_size": 50,216        "use_async_envs": false217    },218    "wandb": {219        "enable": false,220        "disable_artifact": false,221        "project": "lerobot",222        "entity": null,223        "notes": null,224        "run_id": null,225        "mode": null226    },227    "peft": null,228    "use_rabc": false,229    "rabc_progress_path": null,230    "rabc_kappa": 0.01,231    "rabc_epsilon": 1e-06,232    "rabc_head_mode": "sparse",233    "rename_map": {234        "observation.images.top": "observation.images.camera1",235        "observation.images.wrist": "observation.images.camera2",236        "observation.images.side": "observation.images.camera3"237    },238    "checkpoint_path": null239}