Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

sil loss problem #4

Open
MMMMichaelzhang opened this issue Jul 3, 2022 · 2 comments
Open

sil loss problem #4

MMMMichaelzhang opened this issue Jul 3, 2022 · 2 comments

Comments

@MMMMichaelzhang
Copy link

Screenshot from 2022-06-19 22-30-27
Screenshot from 2022-07-03 13-01-58
I trained 3 times with different data, the data includes talking and singing, but the eval sil loss goes up each time. Can you help me with this problem? thank you very much. @yl4579

@skol101
Copy link

skol101 commented Jul 8, 2022

Have you tried using data_augmentation flag?

@yl4579
Copy link
Owner

yl4579 commented Aug 22, 2022

Sorry for the late reply. I was pretty busy recently. This is likely due to some miscongfiguration here. The proportion between F0 and silence loss is not well balanced so the silence loss starts to overfit. You may try to set lambda_f0 = 0.01 or even smaller so they are on the same scale, or use a very big dataset like LibriTTS (the dataset that the pre-trained model was trained on).

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants