search

auspicious3000 / SpeechSplit

Unsupervised Speech Decomposition Via Triple Information Bottleneck
http://arxiv.org/abs/2004.11284
MIT License
636 stars 92 forks source link

Learning rate for P (F0_Converter) #47

Closed Merlin-721 closed 3 years ago

Merlin-721 commented 3 years ago

I have integrated training for P in solver.py but am unsure of what learning rate to use.

The default for G is 0.001, but I doubt this is also correct for P.

What initial LR should I use for P?

Many thanks.

auspicious3000 commented 3 years ago

The default for G should be 0.0001 instead of 0.001. The same works for P as well, but feel free to try other lrs.

Merlin-721 commented 3 years ago

My mistake - 0.0001 is correct. Many thanks for your fast response.