vctk_model_conf.json

{
  "character_embedding_dim": 256,
  "max_text_sequence_length": 80,
  "num_speakers": 108,
  "speaker_embedding_dim": 16,
  "downsample": true,
  "max_duration_secs": 4.0,
  "sample_rate": 16000,
  "mel_bands": 80,
  "n_frames_per_pred": 1,
  "n_fft": 1024,
  "fmin": 125,
  "fmax": 7600,
  "hop_length": 600,
  "win_length": 1024,
  "encoder_channels": 128,
  "num_encoder_conv_blocks": 7,
  "decoder_channels": 256,
  "decoder_prenet_sizes": [128, 256],
  "num_decoder_conv_blocks": 4,
  "decoder_attention_flags": [true, false, false, true],
  "attention_hidden_size": 256,
  "query_position_rate": 1.0,
  "key_position_rate": 1.28, 
  "use_guided_attention": true,
  "guided_attention_g": 0.2,
  "num_converter_conv_blocks": 6,
  "converter_channels": 256,
  "dropout_rate": 0.05,
  "init_type": "xavier",
  "embed_normal_init_weights_std_dev": 0.01,
  "layer_normal_init_weights_std_dev": 0.01,
  "data_shuffling_seed": 3024,
  "numpy_seed": 0,
  "inference_look_ahead": 3,
  "power": 1.4,
  "amp_factor": 5.0
}