-
Notifications
You must be signed in to change notification settings - Fork 21.4k
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
Enable bfloat16 for hardtanh_backward_cuda #91511
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/91511
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit b5ab130: This comment was automatically generated by Dr. CI and updates every 15 minutes. |
|
cc @rohithkrn - just checking, do you recall if there was a reason to not add hardtanh_backward, like some test not passing? or just an oversight in this: #32065 |
@pytorchbot rebase |
You don't have permissions to rebase this PR, only people with write permissions may rebase PRs. |
@cchan honestly don't remember. It's been a while. |
@pytorchbot merge |
Merge startedYour change will be merged once all checks pass (ETA 0-4 Hours). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
Merge failedReason: This PR is too stale; the last push date was more than 3 days ago. Please rebase and try again. You can rebase by leaving the following comment on this PR: Details for Dev Infra teamRaised by workflow job |
@pytorchbot rebase |
You don't have permissions to rebase this PR, only people with write permissions may rebase PRs. |
@pytorchbot rebase |
@pytorchbot successfully started a rebase job. Check the current status here |
Successfully rebased |
a89729c
to
b5ab130
Compare
@pytorchbot merge |
Merge startedYour change will be merged once all checks pass (ETA 0-4 Hours). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
I'm not sure why this was left out in the first place as all adjacent operations have both Half and BFloat16. Things seem to work as expected and this enables
relu6
to be used in bfloat16 training. Hardtanh backward is super simple and precision is not relevant.Previously would fail with: