mirror of
https://github.com/storytold/storyteller-ml.git
synced 2026-10-24 08:36:10 +00:00
ZDisket
456fc04e63
- Tacotron2 based on https://github.com/keonlee9420/Comprehensive-Tacotron2 - Removed DeepSpeaker and Tensorflow dependency - Added ARPA function - Replaced guided attention loss with Convolutional Attention Consistency
29 lines
664 B
YAML
29 lines
664 B
YAML
dataset: "LJSpeech"
|
|
|
|
path:
|
|
corpus_path: "./datasets"
|
|
preprocessed_path: "./preprocessed_data/SingleSpk"
|
|
|
|
preprocessing:
|
|
sanity_check: False # if Ture, then preprocess a single data and quit
|
|
val_size: 512
|
|
text:
|
|
text_cleaners: ["vbasic_cleaners"]
|
|
language: "en"
|
|
audio:
|
|
sampling_rate: 44100
|
|
max_wav_value: 32768.0
|
|
skip_len: 14848
|
|
trim_top_db: 23
|
|
silence_audio_size: 5
|
|
pre_emphasis: False
|
|
stft:
|
|
filter_length: 2048
|
|
hop_length: 512
|
|
win_length: 2048
|
|
mel:
|
|
n_mel_channels: 80
|
|
mel_fmin: 20
|
|
mel_fmax: 11025 # please set to 8000 for HiFi-GAN vocoder, set to null for MelGAN vocoder
|
|
normalize: False
|