{
    "model_type": "diffrhythm",
    "model": {
        "dim": 2048,
        "depth": 16,
        "heads": 32,
        "ff_mult": 4,
        "text_dim": 512,
        "conv_layers": 4,
        "mel_dim": 64,
        "text_num_embeds": 363
    }
}