encoder:
  encoder_kernel_size: 5
  encoder_n_convolutions: 3
  encoder_embedding_dim: 512
  symbols_embedding_dim: 512
decoder:
  n_frames_per_step: 1 # any of reduction factor is supported
  decoder_rnn_dim: 1024
  prenet_dim: 256
  max_decoder_steps: 1000
  gate_threshold: 0.5
  p_attention_dropout: 0.1
  p_decoder_dropout: 0.1
attention:
  attention_rnn_dim: 1024
  attention_dim: 128
location_layer:
  attention_location_n_filters: 32
  attention_location_kernel_size: 31
postnet:
  postnet_embedding_dim: 512
  postnet_kernel_size: 5
  postnet_n_convolutions: 5

# gst:
#   use_gst: False
#   conv_filters: [32, 32, 64, 64, 128, 128]
#   gru_hidden: 128
#   token_size: 128
#   n_style_token: 10
#   attn_head: 4

multi_speaker: False

max_seq_len: 1000

vocoder:
  model: "HiFi-GAN" # support 'HiFi-GAN', 'MelGAN'
  speaker: "LJSpeech" # support  'LJSpeech', 'universal'