| { |
| "type": "vila_chunk", |
| "n_obs_steps": 2, |
| "input_features": { |
| "observation.state": { |
| "type": "STATE", |
| "shape": [ |
| 7 |
| ] |
| }, |
| "observation.images.top": { |
| "type": "VISUAL", |
| "shape": [ |
| 3, |
| 384, |
| 384 |
| ] |
| }, |
| "observation.images.up": { |
| "type": "VISUAL", |
| "shape": [ |
| 3, |
| 384, |
| 384 |
| ] |
| }, |
| "observation.images.side": { |
| "type": "VISUAL", |
| "shape": [ |
| 3, |
| 384, |
| 384 |
| ] |
| }, |
| "observation.images.wrist": { |
| "type": "VISUAL", |
| "shape": [ |
| 3, |
| 384, |
| 384 |
| ] |
| }, |
| "observation.images.image": { |
| "type": "VISUAL", |
| "shape": [ |
| 3, |
| 384, |
| 384 |
| ] |
| }, |
| "observation.image": { |
| "type": "VISUAL", |
| "shape": [ |
| 3, |
| 384, |
| 384 |
| ] |
| }, |
| "observation.images.exterior_image_1_left": { |
| "type": "VISUAL", |
| "shape": [ |
| 3, |
| 384, |
| 384 |
| ] |
| }, |
| "observation.images.exterior_1_left": { |
| "type": "VISUAL", |
| "shape": [ |
| 3, |
| 384, |
| 384 |
| ] |
| }, |
| "observation.images.wrist_left": { |
| "type": "VISUAL", |
| "shape": [ |
| 3, |
| 384, |
| 384 |
| ] |
| } |
| }, |
| "output_features": { |
| "action": { |
| "type": "ACTION", |
| "shape": [ |
| 8, |
| 7 |
| ] |
| } |
| }, |
| "device": "cuda", |
| "use_amp": false, |
| "push_to_hub": true, |
| "repo_id": "FT-LLM-2026-RAMEN/v4-8b-decay2m-ipt_v3_1-instruct4-add-tokens-nvila-stage3-vla-flow-taskfix-20260302-1717", |
| "private": null, |
| "tags": null, |
| "license": null, |
| "pretrained_path": null, |
| "chunk_size": 8, |
| "n_action_steps": 4, |
| "normalization_mapping": { |
| "VISUAL": "IDENTITY", |
| "STATE": "MEAN_STD", |
| "ACTION": "MEAN_STD" |
| }, |
| "temporal_ensemble_mode": "exponential", |
| "temporal_ensemble_decay": 0.1, |
| "temporal_ensemble_every_step": true, |
| "max_state_dim": 14, |
| "max_action_dim": 14, |
| "state_dim": 7, |
| "action_dim": 7, |
| "resize_imgs_with_padding": [ |
| 448, |
| 448 |
| ], |
| "preprocess_images_in_policy": true, |
| "image_key": null, |
| "image_keys": [ |
| "observation.images.top", |
| "observation.images.up", |
| "observation.images.side", |
| "observation.images.wrist", |
| "observation.images.image", |
| "observation.image", |
| "observation.images.exterior_image_1_left", |
| "observation.images.exterior_1_left", |
| "observation.images.wrist_left" |
| ], |
| "auto_detect_image_key": true, |
| "vila_model_name": "FT-LLM-2026-RAMEN/v4-8b-decay2m-ipt_v3_1-instruct4-add-tokens-nvila-stage3", |
| "llm_jp_vila_path": "/home/suzuki/ft_llm_2026_ramen/vla/lerobot_policy_vila/../eval/simpler_env/a100/vla_inference/llm-jp-VILA", |
| "llm_hidden_size": 4096, |
| "hidden_state_layer": -1, |
| "load_in_4bit": false, |
| "bnb_4bit_quant_type": "nf4", |
| "bnb_4bit_use_double_quant": true, |
| "use_flash_attention": true, |
| "action_head_type": "flow", |
| "action_head_hidden_dim": 256, |
| "action_head_num_layers": 2, |
| "action_head_dropout": 0.1, |
| "input_scale": 0.015625, |
| "policy_d_model": 1024, |
| "vision_target_grid": 12, |
| "resampler_num_latents": 32, |
| "resampler_layers": 2, |
| "history_mixer": "gru", |
| "text_encoder_layers": 2, |
| "flow_steps": 10, |
| "flow_depth": 8, |
| "flow_nhead": 8, |
| "flow_dropout": 0.1, |
| "flow_noise_scale": 1.0, |
| "use_state_concat": true, |
| "optimizer_lr": 0.0001, |
| "optimizer_betas": [ |
| 0.9, |
| 0.95 |
| ], |
| "optimizer_eps": 1e-08, |
| "optimizer_weight_decay": 1e-10, |
| "optimizer_grad_clip_norm": 10.0, |
| "use_lr_scheduler": true, |
| "num_warmup_steps": 500, |
| "num_decay_steps": 29500, |
| "decay_lr": 1e-06, |
| "freeze_vision_encoder": true, |
| "freeze_llm_backbone": true, |
| "train_action_head": true, |
| "train_state_proj": true, |
| "vision_unfreeze_last_n_blocks": 0, |
| "vision_lora_r": 0, |
| "vision_lora_alpha": 16, |
| "vision_lora_dropout": 0.0, |
| "vision_lora_target_modules": [ |
| "q_proj", |
| "k_proj", |
| "v_proj", |
| "out_proj", |
| "fc1", |
| "fc2" |
| ], |
| "action_clip_min": null, |
| "action_clip_max": null, |
| "simplerenv_gripper_conversion": true, |
| "simplerenv_gripper_index": -1, |
| "simplerenv_gripper_invert": false, |
| "simplerenv_gripper_binarize": true, |
| "simplerenv_euler_to_axis_angle": true, |
| "simplerenv_rotation_start_index": 3, |
| "prompt_template": "<image>\n\u30ed\u30dc\u30c3\u30c8\u306e\u52d5\u4f5c\u3092\u4e88\u6e2c\u3057\u3066\u304f\u3060\u3055\u3044\u3002", |
| "conversation_template": "llmjp_v3" |
| } |
