config.yml

log_dir: "Models/run_timbre_norm_ctc_titanet/"
save_freq: 1
log_interval: 10
save_interval: 1000
device: "cuda"
epochs: 1000 # number of epochs for first stage training (pre-training)
batch_size: 4
batch_length: 100 # maximum duration of audio in a batch (in seconds)
max_len: 80 # maximum number of frames
pretrained_model: ""
load_only_params: False # set to true if do not want to load epoch numbers and optimizer parameters

F0_path: "modules/JDC/bst.t7"

data_params:
  train_data: "./data/train.txt"
  val_data: "./data/val.txt"
  root_path: "./data/"

preprocess_params:
  sr: 24000
  spect_params:
    n_fft: 2048
    win_length: 1200
    hop_length: 300

model_params:
  fixed: True
  causal: True
  lstm: 2
  norm_f0: True
  use_gr_content_f0: False
  use_gr_prosody_phone: False
  use_gr_timbre_prosody: False
  separate_prosody_encoder: True
  n_c_codebooks: 2
  timbre_norm: True
  use_gr_content_global_f0: True
  w2v: 'w2v-ctc'

  DAC:
    encoder_dim: 64
    encoder_rates: [2, 5, 5, 6]
    decoder_dim: 1536
    decoder_rates: [ 6, 5, 5, 2 ]
    sr: 24000

loss_params:
  base_lr: 0.0001
  discriminator_iter_start: 2000
  lambda_spk: 1.0
  lambda_mel: 45
  lambda_f0: 1.0
  lambda_uv: 1.0