| { |
| "type": "wam", |
| "n_obs_steps": 1, |
| "input_features": { |
| "observation.state": { |
| "type": "STATE", |
| "shape": [ |
| 6 |
| ] |
| }, |
| "observation.images.top": { |
| "type": "VISUAL", |
| "shape": [ |
| 3, |
| 480, |
| 640 |
| ] |
| }, |
| "observation.images.wrist": { |
| "type": "VISUAL", |
| "shape": [ |
| 3, |
| 480, |
| 640 |
| ] |
| } |
| }, |
| "output_features": { |
| "action": { |
| "type": "ACTION", |
| "shape": [ |
| 6 |
| ] |
| } |
| }, |
| "device": "cuda", |
| "use_amp": true, |
| "use_peft": false, |
| "push_to_hub": true, |
| "repo_id": "maximellerbach/wam_multiview", |
| "private": null, |
| "tags": null, |
| "license": null, |
| "pretrained_path": null, |
| "n_obs_frames": 1, |
| "obs_frame_skip": 5, |
| "pred_frame_skip": 30, |
| "n_pred_frames": 1, |
| "image_size": [ |
| 256, |
| 256 |
| ], |
| "horizon": 30, |
| "n_action_steps": 30, |
| "normalization_mapping": { |
| "VISUAL": "MEAN_STD", |
| "STATE": "MEAN_STD", |
| "ACTION": "MEAN_STD" |
| }, |
| "hidden_dim": 384, |
| "pre_norm": false, |
| "vision_backbone": "resnet18", |
| "pretrained_backbone_weights": "ResNet18_Weights.IMAGENET1K_V1", |
| "use_pretrained_backbone": true, |
| "freeze_backbone": false, |
| "ema_decay": 0.99, |
| "latent_dim": 32, |
| "dim_feedforward": 512, |
| "n_vae_encoder_layers": 4, |
| "vae_num_heads": 8, |
| "dropout": 0.1, |
| "action_head_dropout": 0.1, |
| "predictor_num_layers": 12, |
| "predictor_num_heads": 12, |
| "predictor_dropout": 0.05, |
| "predictor_max_tokens": 512, |
| "optimizer_lr": 5e-05, |
| "optimizer_weight_decay": 0.01, |
| "optimizer_grad_clip_norm": 1.0, |
| "scheduler_decay_lr": 1e-05, |
| "scheduler_warmup_steps": 500, |
| "scheduler_decay_steps": 5000 |
| } |
