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

Minor bug fix: changing train_step in examples code to take a mean of the stats instead of taking from the first device. Because the optimizer syncs its own stats (like loss), this didn't matter except for stats returned from the kfac_jax optimizer (or Optax optimizers using OptaxWrapper). However, the Polyak averaged loss wasn't actually synced across devices (as its not part of the optimizer anymore), so "loss_polyak" was being reported only for the first device. #304

Closed

copybara-service wants to merge 0 commits into main from test_700710272

test_700710272% was force-pushed and no longer has any new commits.

Pushing new commits will allow the pull request to be re-opened.