hparams.json

{
  "model_type": "acoustic",
  "in_channels": 297,
  "target_channels": 1028,
  "mgc_target_channels": 67,
  "phoneme_in_channels": 226,
  "emotional_prosodic_in_channels": 71,
  "duration_units": 256,
  "duration_convolutional_bank_widths": [1, 2, 3, 4, 5, 6, 7, 8],
  "duration_max_pooling_width": 2,
  "duration_highway_layers": 1,
  "duration_gru_layer": 2,
  "duration_hidden_size": 32,
  "duration_target_channels": 1,
  "duration_input_padded": 0,
  "duration_target_padded": 0,
  "acoustic_units": 128,
  "acoustic_convolutional_bank_widths": [5, 10, 15, 20, 25, 30, 35, 40],
  "acoustic_max_pooling_width": 10,
  "acoustic_highway_layers": 2,
  "acoustic_gru_layer": 2,
  "acoustic_input_padded": 0,
  "acoustic_target_padded": 0,
  "spec_hidden_size": 128,
  "mgc_hidden_size": 128,
  "energy_hidden_size": 16,
  "cap_hidden_size": 16,
  "bap_hidden_size": 16,
  "lf0_hidden_size": 32,
  "spec_units": 512,
  "mgc_units": 60,
  "energy_units": 1,
  "cap_units": 513,
  "bap_units": 5,
  "lf0_units": 1,
  "uv_units": 2,
  "batch_size": 16,
  "dropout_rate": 0.5,
  "initial_lr": 1e-3,
  "min_lr": 1e-5,
  "weight_decay": 5e-5,
  "max_epochs": 300,
  "gpu_ids": "0"
}