{
    "embed_dim": 1024,
    "vision_cfg": {
        "image_size": 448,
        "layers": [
            3,
            15,
            36,
            10
        ],
        "width": 128,
        "patch_size": null
    },
    "text_cfg": {
        "context_length": 77,
        "vocab_size": 49408,
        "width": 1024,
        "heads": 16,
        "layers": 12
    }
}