{
    "type": "sarm",
    "n_obs_steps": 8,
    "input_features": {
        "observation.images.base": {
            "type": "VISUAL",
            "shape": [
                480,
                640,
                3
            ]
        },
        "observation.state": {
            "type": "STATE",
            "shape": [
                32
            ]
        }
    },
    "output_features": {
        "action": {
            "type": "ACTION",
            "shape": [
                16
            ]
        }
    },
    "device": "cuda",
    "use_amp": false,
    "push_to_hub": true,
    "repo_id": "pepijn223/sarm_1k",
    "private": null,
    "tags": null,
    "license": null,
    "pretrained_path": null,
    "annotation_mode": "dual",
    "frame_gap": 30,
    "max_rewind_steps": 4,
    "image_dim": 512,
    "text_dim": 512,
    "hidden_dim": 768,
    "num_heads": 12,
    "num_layers": 8,
    "max_state_dim": 32,
    "drop_n_last_frames": 1,
    "batch_size": 64,
    "clip_batch_size": 64,
    "dropout": 0.1,
    "stage_loss_weight": 1.0,
    "rewind_probability": 0.8,
    "language_perturbation_probability": 0.2,
    "num_sparse_stages": 1,
    "sparse_subtask_names": [
        "Fold t-shirt"
    ],
    "sparse_temporal_proportions": [
        1.0
    ],
    "num_dense_stages": 5,
    "dense_subtask_names": [
        "Do first fold with both arms",
        "Slide t-shirt closer",
        "Do second fold with both arms",
        "Do third fold with one arm",
        "Do final fold with one arm to make a nice square folded t-shirt"
    ],
    "dense_temporal_proportions": [
        0.14711259363076604,
        0.08305116624302009,
        0.14677170180966917,
        0.16127991987952986,
        0.46178461843701474
    ],
    "pretrained_model_path": null,
    "image_key": "observation.images.base",
    "state_key": "observation.state",
    "normalization_mapping": {
        "VISUAL": "IDENTITY",
        "STATE": "MEAN_STD",
        "LANGUAGE": "IDENTITY",
        "REWARD": "IDENTITY"
    }
}