config/vitssvc.json

{
    "base_config": "config/base.json",
    "model_type": "VITS",
    "task_type": "svc",
    "preprocess": {
        "extract_phone": false,
        "extract_mel": true,
        "extract_linear_spec": true,
        "extract_audio": true,
        "use_linear": true,
        "use_mel": true,
        "use_audio": true,
        "use_text": false,
        "use_phone": true,
        
        "fmin": 0,
        "fmax": null,
        "f0_min": 50,
        "f0_max": 1100,
        // f0_bin in sovits
        "pitch_bin": 256,
        // filter_length in sovits
        "n_fft": 2048,
        // hop_length in sovits
        "hop_size": 512,
        // win_length in sovits
        "win_size": 2048,
        "segment_size": 8192,
        "n_mel": 100,
        "sample_rate": 44100,

        "mel_min_max_stats_dir": "mel_min_max_stats",
        "whisper_dir": "whisper",
        "contentvec_dir": "contentvec",
        "wenet_dir": "wenet",
        "mert_dir": "mert",
    },
    "model": {
        "condition_encoder": {
            "merge_mode": "add",
            "input_melody_dim": 1,
            "use_log_f0": true,
            "n_bins_melody": 256,
            //# Quantization (0 for not quantization)
            "output_melody_dim": 196,
            "input_loudness_dim": 1,
            "use_log_loudness": false,
            "n_bins_loudness": 256,
            "output_loudness_dim": 196,
            "use_whisper": false,
            "use_contentvec": false,
            "use_wenet": false,
            "use_mert": false,
            "whisper_dim": 1024,
            "contentvec_dim": 256,
            "mert_dim": 256,
            "wenet_dim": 512,
            "content_encoder_dim": 196,
            "output_singer_dim": 196,
            "singer_table_size": 512,
            "output_content_dim": 196,
            "use_spkid": true
        },
        "vits": {
            "filter_channels": 256,
            "gin_channels": 256,
            "hidden_channels": 192,
            "inter_channels": 192,
            "kernel_size": 3,
            "n_flow_layer": 4,
            "n_heads": 2,
            "n_layers": 6,
            "n_layers_q": 3,
            "n_speakers": 512,
            "p_dropout": 0.1,
            "ssl_dim": 256,
            "use_spectral_norm": false,
        },
        "generator": "hifigan",
        "generator_config": {
            "hifigan": {
                "resblock": "1",
                "resblock_kernel_sizes": [
                    3,
                    7,
                    11
                ],
                "upsample_rates": [
                    8,8,2,2,2
                ],
                "upsample_kernel_sizes": [
                    16,16,4,4,4
                ],
                "upsample_initial_channel": 512,
                "resblock_dilation_sizes": [
                    [1,3,5],
                    [1,3,5],
                    [1,3,5]
                ]
            },
            "melgan": {
                "ratios": [8, 8, 2, 2, 2],
                "ngf": 32,
                "n_residual_layers": 3,
                "num_D": 3,
                "ndf": 16,
                "n_layers": 4,
                "downsampling_factor": 4
            },
            "bigvgan": {
                "resblock": "1",
                "activation": "snakebeta",
                "snake_logscale": true,
                "upsample_rates": [
                    8,8,2,2,2,
                ],
                "upsample_kernel_sizes": [
                    16,16,4,4,4,
                ],
                "upsample_initial_channel": 512,
                "resblock_kernel_sizes": [
                    3,
                    7,
                    11
                ],
                "resblock_dilation_sizes": [
                    [1,3,5],
                    [1,3,5],
                    [1,3,5]
                ]
            },
            "nsfhifigan": {
                "resblock": "1",
                "harmonic_num": 8,
                "upsample_rates": [
                    8,8,2,2,2,
                ],
                "upsample_kernel_sizes": [
                    16,16,4,4,4,
                ],
                "upsample_initial_channel": 768,
                "resblock_kernel_sizes": [
                    3,
                    7,
                    11
                ],
                "resblock_dilation_sizes": [
                    [1,3,5],
                    [1,3,5],
                    [1,3,5]
                ]
            },
            "apnet": {
              "ASP_channel": 512,
              "ASP_resblock_kernel_sizes": [3,7,11],
              "ASP_resblock_dilation_sizes": [[1,3,5], [1,3,5], [1,3,5]],
              "ASP_input_conv_kernel_size": 7,
              "ASP_output_conv_kernel_size": 7,
        
              "PSP_channel": 512,
              "PSP_resblock_kernel_sizes": [3,7,11],
              "PSP_resblock_dilation_sizes": [[1,3,5], [1,3,5], [1,3,5]], 
              "PSP_input_conv_kernel_size": 7,
              "PSP_output_R_conv_kernel_size": 7,
              "PSP_output_I_conv_kernel_size": 7,
            }
        },
    },
    "train": {
        "fp16_run": true,
        "learning_rate": 2e-4,
        "betas": [
            0.8,
            0.99
        ],
        "eps": 1e-9,
        "batch_size": 16,
        "lr_decay": 0.999875,
        // "segment_size": 8192,
        "init_lr_ratio": 1,
        "warmup_epochs": 0,
        "c_mel": 45,
        "c_kl": 1.0,
        "AdamW": {
            "betas": [
                0.8,
                0.99
            ],
            "eps": 1e-9,
        }
    }
}