dinglx/smolvla_stacking_model

1{2    "type": "smolvla",3    "n_obs_steps": 1,4    "normalization_mapping": {5        "VISUAL": "IDENTITY",6        "STATE": "MEAN_STD",7        "ACTION": "MEAN_STD"8    },9    "input_features": {10        "observation.state": {11            "type": "STATE",12            "shape": [13                614            ]15        },16        "observation.images.top": {17            "type": "VISUAL",18            "shape": [19                3,20                480,21                64022            ]23        },24        "observation.images.wrist": {25            "type": "VISUAL",26            "shape": [27                3,28                480,29                64030            ]31        }32    },33    "output_features": {34        "action": {35            "type": "ACTION",36            "shape": [37                638            ]39        }40    },41    "device": "cuda",42    "use_amp": false,43    "push_to_hub": true,44    "repo_id": "dinglx/smolvla_stacking_model",45    "private": null,46    "tags": null,47    "license": null,48    "chunk_size": 50,49    "n_action_steps": 50,50    "max_state_dim": 32,51    "max_action_dim": 32,52    "resize_imgs_with_padding": [53        512,54        51255    ],56    "empty_cameras": 0,57    "adapt_to_pi_aloha": false,58    "use_delta_joint_actions_aloha": false,59    "tokenizer_max_length": 48,60    "num_steps": 10,61    "use_cache": true,62    "freeze_vision_encoder": true,63    "train_expert_only": true,64    "train_state_proj": true,65    "optimizer_lr": 0.0001,66    "optimizer_betas": [67        0.9,68        0.9569    ],70    "optimizer_eps": 1e-08,71    "optimizer_weight_decay": 1e-10,72    "optimizer_grad_clip_norm": 10.0,73    "scheduler_warmup_steps": 1000,74    "scheduler_decay_steps": 30000,75    "scheduler_decay_lr": 2.5e-06,76    "vlm_model_name": "HuggingFaceTB/SmolVLM2-500M-Video-Instruct",77    "load_vlm_weights": true,78    "add_image_special_tokens": false,79    "attention_mode": "cross_attn",80    "prefix_length": 0,81    "pad_language_to": "max_length",82    "num_expert_layers": 0,83    "num_vlm_layers": 16,84    "self_attn_every_n_layers": 2,85    "expert_width_multiplier": 0.75,86    "min_period": 0.004,87    "max_period": 4.088}