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.

Sign up for GitHub

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

Jump to bottom

Train loss NaN when previous QA is set to 0 #8

Closed
Ajanulk opened this issue Mar 16, 2019 · 2 comments
Closed

Train loss NaN when previous QA is set to 0 #8

Ajanulk opened this issue Mar 16, 2019 · 2 comments

Comments

@Ajanulk
Copy link

Ajanulk commented Mar 16, 2019

When I set previous QA to 0 in conf, the train loss met nan. How can I solve it?
@zcgzcgzcg1
Copy link
Contributor

It may happen that certain weights get very large. Try using smaller gradient clipping (like 1, 0.1).

@Ajanulk
Copy link
Author

Ajanulk commented Apr 2, 2019

It may happen that certain weights get very large. Try using smaller gradient clipping (like 1, 0.1).

@zcgzcgzcg1 Thanks for your reply. But it doesnt work according to your suggestion (gradient clipping 1, 0.1).

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
2 participants
@zcgzcgzcg1 @Ajanulk