{
    "model_type": "FlowMatchingTransformer",
    "dataset": [
        "emilia"
    ],
    "preprocess": {
        "hop_size": 480,
        "sample_rate": 24000,
        "n_fft": 1920,
        "num_mels": 128,
        "win_size": 1920,
        "fmin": 0,
        "fmax": 12000,
        "mel_var": 8.14,
        "mel_mean": -4.92,
        "processed_dir": "",
        "valid_file": "valid.json",
        "train_file": "train.json",
        "use_phone_cond": false,
        "use_emilia_101k": true
    },
    "model": {
        "flow_matching_transformer": {
            "mel_dim": 128,
            "hidden_size": 1024,
            "num_layers": 16,
            "num_heads": 16,
            "cfg_scale": 0.2,
            "use_cond_code": true, // false means Hidden features
            "cond_dim": 1024, // HuBERT features dimension
            "cond_codebook_size": 8192, // VQ Codebook Size
            "cond_scale_factor": 1, // 1 means not use ReTrans
            "use_pretrained_model": false,
            "sigma": 1e-5,
            "time_scheduler": "cos"
        },
        "cond_type": "hubert_codec",
        "cond_sample_rate": 16000,
        "representation_stat_mean_var_path": "./Amphion/models/vc/vevo/config/hubert_large_l18_mean_std.npz",
        "repcodec": {
            "codebook_size": 8192, // VQ Codebook Size
            "hidden_size": 1024, // Representations Dim
            "codebook_dim": 8,
            "vocos_dim": 384,
            "vocos_intermediate_dim": 2048,
            "vocos_num_layers": 12,
        }
    },
}