rishikksh20 / FastSpeech2

PyTorch Implementation of FastSpeech 2 : Fast and High-Quality End-to-End Text to Speech
Apache License 2.0
225 stars 51 forks source link

Question about duration predictor. #29

Closed Jackson-Kang closed 3 years ago

Jackson-Kang commented 3 years ago

Hello @rishikksh20 ,

First of all, thank you for sharing your awesome works such as FastSpeech2, VocGAN, AdaSpeech and etcs. It helped me a lot.

I leave this question due to my curiosity about duration predictor improvements. A few months ago, evaluation performance of duration predictor seemed to be not good due to overfitting. (train error: below 0.01, but eval error: 0.5~0.6.)

But, now it has been drastically improved (eval error 0.5~0.6 -> 0.06 ~ 0.08).

If you don't mind, could you tell me what was the problem of your previous version of duration predictor?

Always appreciate,

rishikksh20 commented 3 years ago

@Jackson-Kang I haven't updated this code repo for a while, which duration predictor you are talking about.

Jackson-Kang commented 3 years ago

I think I just confused with other repos.

Anyway, thank you for answering my question. Sincerely,