{
    "type": "wam",
    "n_obs_steps": 1,
    "input_features": {
        "observation.images.top": {
            "type": "VISUAL",
            "shape": [
                3,
                480,
                640
            ]
        },
        "observation.state": {
            "type": "STATE",
            "shape": [
                14
            ]
        }
    },
    "output_features": {
        "action": {
            "type": "ACTION",
            "shape": [
                14
            ]
        }
    },
    "device": "cuda",
    "use_amp": true,
    "use_peft": false,
    "push_to_hub": true,
    "repo_id": "maximellerbach/test_wam_decoder",
    "private": null,
    "tags": null,
    "license": null,
    "pretrained_path": "outputs/train/wam_aloha_decode_test/checkpoints/last/pretrained_model",
    "time_between_frames": 50,
    "lag_offset": 3,
    "image_size": [
        256,
        256
    ],
    "horizon": 50,
    "n_action_steps": 50,
    "normalization_mapping": {
        "VISUAL": "IDENTITY",
        "STATE": "MEAN_STD",
        "ACTION": "MEAN_STD"
    },
    "hidden_dim": 512,
    "dim_feedforward": 2048,
    "pre_norm": false,
    "dropout": 0.1,
    "vision_backbone": "resnet18",
    "pretrained_backbone_weights": "ResNet18_Weights.IMAGENET1K_V1",
    "use_pretrained_backbone": true,
    "freeze_backbone": false,
    "latent_dim": 32,
    "n_vae_encoder_layers": 4,
    "action_head_num_layers": 4,
    "action_head_num_heads": 8,
    "action_head_dropout": 0.1,
    "predictor_num_layers": 8,
    "predictor_num_heads": 8,
    "predictor_dropout": 0.05,
    "predictor_max_tokens": 1024,
    "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
}