{
    "architectures": [
        "SopranoForCausalLM"
    ],
    "bos_token_id": 1,
    "decoder_dim": 512,
    "decoder_intermediate_dim": 1536,
    "decoder_num_layers": 8,
    "dw_kernel": 3,
    "eos_token_id": 2,
    "head_dim": 128,
    "hidden_size": 512,
    "hop_length": 512,
    "intermediate_size": 2304,
    "max_position_embeddings": 512,
    "model_type": "soprano",
    "n_fft": 2048,
    "num_attention_heads": 4,
    "num_hidden_layers": 17,
    "num_key_value_heads": 1,
    "pad_token_id": 0,
    "quantization": {
        "group_size": 64,
        "bits": 4,
        "mode": "affine"
    },
    "quantization_config": {
        "group_size": 64,
        "bits": 4,
        "mode": "affine"
    },
    "receptive_field": 4,
    "rms_norm_eps": 1e-06,
    "rope_theta": 10000,
    "sample_rate": 32000,
    "tie_word_embeddings": false,
    "token_size": 2048,
    "upscale": 4,
    "use_cache": true,
    "vocab_size": 8192
}