model_input: waveform
model_params:
  attn_alpha_init: 0.1
  attn_dim: 256
  attn_dropout: 0.1
  attn_heads: 4
  error_head_dnn_dim: 64
  mean_net_dnn_dim: 64
  mean_net_output_type: scalar
  mean_net_range_clipping: true
  mora_dropout: 0.1
  mora_emb_dim: 256
  mora_ffn_dim: 512
  mora_max_len: 128
  mora_pos_encoding: rope
  mora_transformer_heads: 4
  mora_transformer_layers: 1
  mora_vocab_size: 141
  s3prl_name: wav2vec2
  speaker_grl_dropout: 0.1
  speaker_grl_hidden_dim: 128
  ssl_model_layer_idx: -1
  ssl_model_output_dim: 768
  ssl_module: s3prl
  use_error_head: true
  use_listener_modeling: false
  use_mean_listener: false
  use_mora: true
  use_speaker_grl: true
mora_vocab_path: data/vocab.txt
num_speakers: 13
sampling_rate: 16000