| { |
| "type": "gaze_flow_matching", |
| "n_obs_steps": 1, |
| "input_features": { |
| "observation.images.right": { |
| "type": "VISUAL", |
| "shape": [ |
| 240, |
| 320, |
| 3 |
| ] |
| }, |
| "observation.images.left": { |
| "type": "VISUAL", |
| "shape": [ |
| 240, |
| 320, |
| 3 |
| ] |
| }, |
| "observation.images.user": { |
| "type": "VISUAL", |
| "shape": [ |
| 240, |
| 320, |
| 3 |
| ] |
| } |
| }, |
| "output_features": { |
| "action": { |
| "type": "ACTION", |
| "shape": [ |
| 26 |
| ] |
| } |
| }, |
| "device": "cuda", |
| "use_amp": false, |
| "use_peft": false, |
| "push_to_hub": true, |
| "repo_id": "fecasado/gfm-cubes-baseline2", |
| "private": null, |
| "tags": null, |
| "license": null, |
| "pretrained_path": null, |
| "normalization_mapping": { |
| "VISUAL": "MEAN_STD", |
| "STATE": "MEAN_STD", |
| "ACTION": "MEAN_STD" |
| }, |
| "gaze_feature_dim": 3, |
| "gaze_state_indices": [ |
| -3, |
| -2, |
| -1 |
| ], |
| "gaze_temporal_window": 1, |
| "vision_backbone": "resnet18", |
| "vit_token_size": 16, |
| "vit_foveation_strides": [ |
| 1, |
| 2, |
| 6 |
| ], |
| "vit_foveation_grid_sizes": [ |
| 2, |
| 3, |
| 3 |
| ], |
| "vit_model_size": "base", |
| "vit_pretrained_repo_id": "", |
| "vit_freeze": false, |
| "vit_backbone": "resnet18", |
| "gaze_cross_camera": false, |
| "gaze_cross_camera_method_train": "roma", |
| "gaze_cross_camera_method_inference": "roma", |
| "gaze_cross_camera_match_every": 5, |
| "gaze_cross_camera_timeout": 0.05, |
| "gaze_cross_camera_cache": null, |
| "obs_step_size": 1, |
| "n_action_steps": 2, |
| "horizon": 4, |
| "n_sampling_steps": 32, |
| "use_temporal_ensemble": true, |
| "temporal_ensemble_coeff": -0.01, |
| "dim_model": 512, |
| "n_heads": 8, |
| "mlp_ratio": 4.0, |
| "dropout": 0.1, |
| "dit_n_layers": 8, |
| "dit_time_dim": 256, |
| "pool_n_queries": 16, |
| "pool_n_layers": 4, |
| "use_gaze": false, |
| "use_gaze_as_action": false, |
| "gaze_model_repo_id": "", |
| "gaze_noise": 0.01, |
| "image_norm_mode": "mean_std", |
| "state_norm_mode": "min_max", |
| "action_norm_mode": "min_max", |
| "image_to_gaze_key": {}, |
| "optimizer_lr": 0.0001, |
| "optimizer_lr_backbone": 0.0001, |
| "optimizer_betas": [ |
| 0.95, |
| 0.999 |
| ], |
| "optimizer_eps": 1e-08, |
| "optimizer_weight_decay": 1e-06, |
| "scheduler_name": "cosine", |
| "scheduler_warmup_steps": 500, |
| "use_ema": true, |
| "ema_decay": 0.99, |
| "input_shape": [ |
| 240, |
| 320 |
| ], |
| "visualize": true |
| } |
