config.json

{
    "type": "rewind",
    "n_obs_steps": 1,
    "input_features": {
        "observation.state": {
            "type": "STATE",
            "shape": [
                16
            ]
        },
        "observation.images.left_wrist": {
            "type": "VISUAL",
            "shape": [
                3,
                480,
                640
            ]
        },
        "observation.images.right_wrist": {
            "type": "VISUAL",
            "shape": [
                3,
                480,
                640
            ]
        },
        "observation.images.base": {
            "type": "VISUAL",
            "shape": [
                3,
                480,
                640
            ]
        }
    },
    "output_features": {
        "action": {
            "type": "ACTION",
            "shape": [
                16
            ]
        }
    },
    "device": "cuda",
    "use_amp": false,
    "push_to_hub": true,
    "repo_id": "pepijn223/rewind-3-fold-final",
    "private": null,
    "tags": null,
    "license": null,
    "pretrained_path": null,
    "video_dim": 768,
    "text_dim": 384,
    "hidden_dim": 512,
    "num_heads": 8,
    "num_layers": 4,
    "max_length": 32,
    "subsample_video": true,
    "use_temporal_sampler": true,
    "sequence_stride": 1,
    "batch_size": 64,
    "dino_batch_size": 64,
    "gradient_checkpointing": false,
    "pretrained_model_path": null,
    "dropout": 0.1,
    "image_key": "observation.images.base",
    "task_description": "perform the task",
    "encode_on_the_fly": true,
    "use_dataset_task": true
}