config.json

{
  "train": {
    "log_interval": 50,
    "eval_interval": 1000,
    "seed": 1234,
    "port": 8001,
    "epochs": 10000,
    "learning_rate": 0.0002,
    "betas": [
      0.8,
      0.99
    ],
    "eps": 1e-09,
    "batch_size": 6,
    "accumulation_steps": 1,
    "fp16_run": false,
    "lr_decay": 0.998,
    "segment_size": 10240,
    "init_lr_ratio": 1,
    "warmup_epochs": 0,
    "c_mel": 45,
    "keep_ckpts": 4
  },
  "data": {
    "data_dir": "dataset",
    "dataset_type": "SingDataset",
    "collate_type": "SingCollate",
    "training_filelist": "filelists/train.txt",
    "validation_filelist": "filelists/val.txt",
    "max_wav_value": 32768.0,
    "sampling_rate": 44100,
    "n_fft": 2048,
    "fmin": 0,
    "fmax": 22050,
    "hop_length": 512,
    "win_size": 2048,
    "acoustic_dim": 80,
    "c_dim": 256,
    "min_level_db": -115,
    "ref_level_db": 20,
    "min_db": -115,
    "max_abs_value": 4.0,
    "n_speakers": 200
  },
  "model": {
    "hidden_channels": 192,
    "spk_channels": 192,
    "filter_channels": 768,
    "n_heads": 2,
    "n_layers": 4,
    "kernel_size": 3,
    "p_dropout": 0.1,
    "prior_hidden_channels": 192,
    "prior_filter_channels": 768,
    "prior_n_heads": 2,
    "prior_n_layers": 4,
    "prior_kernel_size": 3,
    "prior_p_dropout": 0.1,
    "resblock": "1",
    "use_spectral_norm": false,
    "resblock_kernel_sizes": [
      3,
      7,
      11
    ],
    "resblock_dilation_sizes": [
      [
        1,
        3,
        5
      ],
      [
        1,
        3,
        5
      ],
      [
        1,
        3,
        5
      ]
    ],
    "upsample_rates": [
      8,
      8,
      4,
      2
    ],
    "upsample_initial_channel": 256,
    "upsample_kernel_sizes": [
      16,
      16,
      8,
      4
    ],
    "n_harmonic": 64,
    "n_bands": 65
  },
  "spk": {
    "timdillon": 0
  }
}