config_hubert_general.yaml

generator:
  name: SoundStream
  config:
    n_filters: 32
    D: 256
    target_bandwidths:
    - 0.5
    - 1
    - 1.5
    - 2
    - 4
    ratios:
    - 8
    - 5
    - 4
    - 2
    sample_rate: 16000
    bins: 1024
    semantic_techer: hubert_base_general
d_list:
- mfd
mfd:
  name: MultiFrequencyDiscriminator
  config:
    hop_lengths:
    - 32
    - 64
    - 128
    - 256
    - 512
    - 1024
    hidden_channels:
    - 64
    - 128
    - 256
    - 512
    - 512
    - 512
    domain: double
    mel_scale: true
    sample_rate: 16000
mpd:
  name: MultiPeriodDiscriminator
  config:
    period_sizes:
    - 2
    - 3
    - 5
    - 7
    - 11
    period_kernel_size: 5
msd:
  name: MultiScaleDiscriminator
  config:
    num_scales: 3
    pool_kernel_size: 4
    pool_stride: 2
optimizer:
  g:
    name: AdamW
    config:
      lr: 0.0002
      betas:
      - 0.8
      - 0.99
      eps: 1.0e-06
  d:
    name: AdamW
    config:
      lr: 0.0002
      betas:
      - 0.8
      - 0.99
      eps: 1.0e-06
lr_scheduler:
  g:
    name: ExponentialLR
    config:
      gamma: 0.999
  d:
    name: ExponentialLR
    config:
      gamma: 0.999
criterion:
  g_criterion:
    name: losses.generator_loss.GeneratorSTFTLoss
    config:
      use_mel_loss: false
      adv_criterion: MSEGLoss
      mel_loss_weight: 45
      use_feature_match: true
      feat_match_loss_weight: 20
      use_full_stft_loss: true
      use_sub_stft_loss: true
      full_stft_loss_weight: 1
      sub_stft_loss_weight: 1
      mel_scale_loss:
        sampling_rate: 16000
        n_fft: 1024
        num_mels: 80
        hop_size: 160
        win_size: 800
        fmin: 0
      full_multi_scale_stft_loss:
        fft_sizes:
        - 512
        - 1024
        - 2048
        win_sizes:
        - 480
        - 960
        - 1200
        hop_sizes:
        - 120
        - 240
        - 300
      sub_multi_scale_stft_loss:
        num_bands: 6
        fft_sizes:
        - 128
        - 256
        - 256
        win_sizes:
        - 80
        - 120
        - 200
        hop_sizes:
        - 20
        - 40
        - 50
  d_criterion:
    name: losses.discriminator_loss.MSEDiscriminatorLoss
    config: null
  commit_loss_weight: 1.0
  codebook_loss_weight: 100
audio_norm_scale: 0.95