DataLabHell/smolvla_so100_stack_cubes_v3

1{2    "dataset": {3        "repo_id": "DataLabHell/so100_stack_cubes_3view",4        "repo_type": "dataset",5        "root": null,6        "episodes": null,7        "image_transforms": {8            "enable": false,9            "max_num_transforms": 3,10            "random_order": false,11            "tfs": {12                "brightness": {13                    "weight": 1.0,14                    "type": "ColorJitter",15                    "kwargs": {16                        "brightness": [17                            0.8,18                            1.219                        ]20                    }21                },22                "contrast": {23                    "weight": 1.0,24                    "type": "ColorJitter",25                    "kwargs": {26                        "contrast": [27                            0.8,28                            1.229                        ]30                    }31                },32                "saturation": {33                    "weight": 1.0,34                    "type": "ColorJitter",35                    "kwargs": {36                        "saturation": [37                            0.5,38                            1.539                        ]40                    }41                },42                "hue": {43                    "weight": 1.0,44                    "type": "ColorJitter",45                    "kwargs": {46                        "hue": [47                            -0.05,48                            0.0549                        ]50                    }51                },52                "sharpness": {53                    "weight": 1.0,54                    "type": "SharpnessJitter",55                    "kwargs": {56                        "sharpness": [57                            0.5,58                            1.559                        ]60                    }61                },62                "affine": {63                    "weight": 1.0,64                    "type": "RandomAffine",65                    "kwargs": {66                        "degrees": [67                            -5.0,68                            5.069                        ],70                        "translate": [71                            0.05,72                            0.0573                        ]74                    }75                }76            }77        },78        "revision": null,79        "use_imagenet_stats": true,80        "video_backend": "pyav",81        "return_uint8": false,82        "depth_output_unit": "mm",83        "streaming": false,84        "eval_split": 0.185    },86    "env": null,87    "policy": {88        "type": "smolvla",89        "n_obs_steps": 1,90        "input_features": {91            "observation.state": {92                "type": "STATE",93                "shape": [94                    695                ]96            },97            "observation.images.camera1": {98                "type": "VISUAL",99                "shape": [100                    3,101                    256,102                    256103                ]104            },105            "observation.images.camera2": {106                "type": "VISUAL",107                "shape": [108                    3,109                    256,110                    256111                ]112            },113            "observation.images.camera3": {114                "type": "VISUAL",115                "shape": [116                    3,117                    256,118                    256119                ]120            }121        },122        "output_features": {123            "action": {124                "type": "ACTION",125                "shape": [126                    6127                ]128            }129        },130        "device": "cuda",131        "use_amp": false,132        "use_peft": false,133        "push_to_hub": true,134        "repo_id": "DataLabHell/smolvla_so100_stack_cubes_v3",135        "private": null,136        "tags": null,137        "license": null,138        "pretrained_path": "lerobot/smolvla_base",139        "pretrained_revision": null,140        "chunk_size": 50,141        "n_action_steps": 50,142        "normalization_mapping": {143            "VISUAL": "IDENTITY",144            "STATE": "MEAN_STD",145            "ACTION": "MEAN_STD"146        },147        "max_state_dim": 32,148        "max_action_dim": 32,149        "resize_imgs_with_padding": [150            512,151            512152        ],153        "empty_cameras": 0,154        "adapt_to_pi_aloha": false,155        "use_delta_joint_actions_aloha": false,156        "tokenizer_max_length": 48,157        "num_steps": 10,158        "use_cache": true,159        "freeze_vision_encoder": false,160        "train_expert_only": false,161        "train_state_proj": true,162        "optimizer_lr": 0.0001,163        "optimizer_betas": [164            0.9,165            0.95166        ],167        "optimizer_eps": 1e-08,168        "optimizer_weight_decay": 1e-10,169        "optimizer_grad_clip_norm": 10.0,170        "scheduler_warmup_steps": 1000,171        "scheduler_decay_steps": 30000,172        "scheduler_decay_lr": 2.5e-06,173        "vlm_model_name": "HuggingFaceTB/SmolVLM2-500M-Video-Instruct",174        "load_vlm_weights": true,175        "add_image_special_tokens": false,176        "attention_mode": "cross_attn",177        "prefix_length": 0,178        "pad_language_to": "max_length",179        "num_expert_layers": 0,180        "num_vlm_layers": 16,181        "self_attn_every_n_layers": 2,182        "expert_width_multiplier": 0.75,183        "min_period": 0.004,184        "max_period": 4.0,185        "rtc_config": null,186        "compile_model": false,187        "compile_mode": "max-autotune"188    },189    "reward_model": null,190    "output_dir": "/tmp/train/smolvla_v3",191    "job_name": "smolvla_v3",192    "resume": false,193    "seed": 1000,194    "cudnn_deterministic": false,195    "num_workers": 4,196    "batch_size": 16,197    "prefetch_factor": 4,198    "persistent_workers": true,199    "dataloader_multiprocessing_context": "spawn",200    "steps": 30000,201    "env_eval_freq": 20000,202    "log_freq": 200,203    "eval_steps": 1000,204    "max_eval_samples": 0,205    "tolerance_s": 0.0001,206    "save_checkpoint": true,207    "save_freq": 2500,208    "use_policy_training_preset": true,209    "optimizer": {210        "type": "adamw",211        "lr": 0.0001,212        "weight_decay": 1e-10,213        "grad_clip_norm": 10.0,214        "betas": [215            0.9,216            0.95217        ],218        "eps": 1e-08219    },220    "scheduler": {221        "type": "cosine_decay_with_warmup",222        "num_warmup_steps": 1000,223        "num_decay_steps": 30000,224        "peak_lr": 0.0001,225        "decay_lr": 2.5e-06226    },227    "eval": {228        "n_episodes": 50,229        "batch_size": 22,230        "use_async_envs": true,231        "recording": false,232        "recording_repo_id": null,233        "recording_private": false234    },235    "wandb": {236        "enable": false,237        "disable_artifact": false,238        "project": "lerobot",239        "entity": null,240        "notes": null,241        "run_id": null,242        "mode": null,243        "add_tags": true244    },245    "peft": null,246    "job": {247        "target": null,248        "image": "huggingface/lerobot-gpu:latest",249        "timeout": "2d",250        "detach": false,251        "tags": []252    },253    "save_checkpoint_to_hub": true,254    "sample_weighting": null,255    "rename_map": {256        "observation.images.front": "observation.images.camera1",257        "observation.images.gripper": "observation.images.camera2",258        "observation.images.right_side": "observation.images.camera3"259    },260    "checkpoint_path": null261}