Tex>> Transformer_TTS>> 返回
项目作者： syoamakase

项目描述 ：
Transformer model for Text-to-Speech (TTS)
高级语言： Python
项目地址: git://github.com/syoamakase/Transformer_TTS.git
创建时间： 2020-07-17T05:20:25Z
项目社区：https://github.com/syoamakase/Transformer_TTS
开源协议：
下载


Transformer_TTS

This repository is Text-to-Speech (TTS) project based on Transformer.
It includes FastSpeech~2.

train

python train.py --hp_file config/hparams_template.py
python train_fastspeech2.py --hp_file config/hparams_template.py

If you want to check a loss curve, tensorboard --logdir <save_dir>/logs

hparams.py

hparams.py is a file to control hyper parameters.

When you use your own dataset, you must adjusttrain_script, test_script, mean_file, var_file, and vocab_size.

test

When you generate a speech, please take an average.
python utils/average_checkpoints.py --backend pytorch --snapshots <save directory>/network.epoch* --out <save directory>/network.average_epoch991-epoch1000 --start 991 --end 1000

python test.py --load_name <model path>

python test_fastspeech2.py --load_name <model path>