train_transformer.yaml

audio:
  num_mels: 80
  n_fft: 2048
  sr: 22050
  preemphasis: 0.97
  hop_length: 275
  win_length: 1102
  power: 1.2
  min_level_db: -100
  ref_level_db: 20
  outputs_per_step: 1


hidden_size: 256
embedding_size: 512


warm_up_step: 4000
grad_clip_thresh: 1.0
batch_size: 32
epochs: 10000
lr: 0.001
save_step: 1000
image_step: 2000
use_gpu: True
use_data_parallel: False
stop_token: False

data_path: ../../../dataset/LJSpeech-1.1
save_path: ./checkpoint
log_dir: ./log
#checkpoint_path: ./checkpoint
#ransformer_step: 97000