config.json

{
    "image_dim": 192,
    "downscale_factor": 8,
    "encoder": {
        "group_count": 4,
        "n_hid": 256,
        "n_blk_per_group": 2,
        "input_channels": 3,
        "vocab_size": 8192
    },
    "decoder": {
        "group_count": 4,
        "n_init": 128,
        "n_hid": 256,
        "n_blk_per_group": 2,
        "output_channels": 3,
        "vocab_size": 8192
    }
}