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

Numeric underflow when loss gets small. #13

Open

dhruvrajan opened this issue Oct 4, 2019 · 0 comments

Assignees

Owner

dhruvrajan commented Oct 4, 2019

When training MNIST for > 14 epochs, there seems to be an underflow problem, resulting in NaNs when the loss becomes ~ 7E-3.

dhruvrajan self-assigned this

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment