SyntaSpeech / egs /egs_bases /tts /fs2_orig.yaml
yerfor's picture
init
22871e7
raw
history blame
290 Bytes
base_config: ./fs.yaml
task_cls: tasks.tts.fs2_orig.FastSpeech2OrigTask
encoder_type: fft
decoder_type: fft
use_energy_embed: false
use_pitch_embed: true
pitch_type: cwt # frame|ph|cwt
binarization_args:
with_f0cwt: true
use_gt_energy: false
cwt_std_scale: 0.8
dropout: 0.1
mel_losses: l1