KEMBAR78
Gaussian nll loss scalar variance support by michael-diggin · Pull Request #138931 · pytorch/pytorch · GitHub
Skip to content

Gaussian nll loss scalar variance support #138931

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
Closed
wants to merge 5 commits into from
Closed

Gaussian nll loss scalar variance support #138931

wants to merge 5 commits into from

Conversation

@michael-diggin
Copy link
Contributor

Fixes #138747

Adds support for variance being a Tensor or a float in gaussian_nll_loss to avoid a cpu-gpu sync point in the loss function, when the variance is a static tensor like <scalar>*torch.ones_like(input)

@pytorch-bot pytorch-bot bot added the release notes: nn release notes category label Oct 25, 2024
@linux-foundation-easycla
Copy link

linux-foundation-easycla bot commented Oct 25, 2024 •
edited
Loading

CLA Signed

The committers listed above are authorized under a signed CLA.

@pytorch-bot
Copy link

pytorch-bot bot commented Oct 25, 2024 •
edited
Loading

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/138931

Note: Links to docs will display an error until the docs builds have been completed.

❗ 1 Active SEVs

There are 1 currently active SEVs. If your PR is affected, please view them below:

✅ You can merge normally! (1 Unrelated Failure)

As of commit 6da0ea9 with merge base de34f58 (image):

FLAKY - The following job failed but was likely due to flakiness present on trunk:

This comment was automatically generated by Dr. CI and updates every 15 minutes.

@soulitzer soulitzer added the triaged This issue has been looked at a team member, and triaged and prioritized into an appropriate module label Oct 28, 2024
mikaylagawarecki
mikaylagawarecki approved these changes Nov 6, 2024
View reviewed changes
Copy link
Contributor

@mikaylagawarecki mikaylagawarecki left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks

@mikaylagawarecki
Copy link
Contributor

@pytorchbot merge -r

@pytorch-bot pytorch-bot bot added the ciflow/trunk Trigger trunk jobs on your pull request label Nov 6, 2024
@mikaylagawarecki mikaylagawarecki added topic: not user facing topic category and removed ciflow/trunk Trigger trunk jobs on your pull request labels Nov 6, 2024
@pytorchmergebot
Copy link
Collaborator

@pytorchbot started a rebase job onto refs/remotes/origin/viable/strict. Check the current status here

@pytorchmergebot
Copy link
Collaborator

Successfully rebased gaussian-nll-loss-scalar-var onto refs/remotes/origin/viable/strict, please pull locally before adding more changes (for example, via git checkout gaussian-nll-loss-scalar-var && git pull --rebase)

@pytorchmergebot pytorchmergebot force-pushed the gaussian-nll-loss-scalar-var branch from 7e35745 to 50ba45c Compare November 6, 2024 23:35
@pytorchmergebot
Copy link
Collaborator

Merge started

Your 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

Advanced Debugging
Check the merge workflow status
here

@pytorchmergebot
Copy link
Collaborator

Merge failed

Reason: 3 mandatory check(s) failed. The first few are:

Dig deeper by viewing the failures on hud

Details for Dev Infra team Raised by workflow job

Failing merge rule: Core Maintainers

@mikaylagawarecki
Copy link
Contributor

Seems like CI is failing

@michael-diggin
Copy link
Contributor Author

Seems like CI is failing

I think it wasn't liking the fact I had used Tensor | float as that isn't supported on Python 3.9. I've changed it to use Union[Tensor, Float] which should be the exact same.

@michael-diggin
Copy link
Contributor Author

@mikaylagawarecki I've updated the PR which should fix the failing CI, if you would be able to re-trigger that workflow if you're happy with the update?

@michael-diggin
Copy link
Contributor Author

Thanks! I think the only remaining failure is the bc-lint here: https://github.com/pytorch/pytorch/actions/runs/11784651031/job/32831020347?pr=138931#step:2:833
Which I believe is a false positive, as Tensor -> Union[Tensor, Float] is a backwards compatible change.
If you agree we can add the suppress-bc-linter label and I think that should raise them as notifications rather than errors?

@michael-diggin
Copy link
Contributor Author

Thanks! I think the only remaining failure is the bc-lint here: https://github.com/pytorch/pytorch/actions/runs/11784651031/job/32831020347?pr=138931#step:2:833 Which I believe is a false positive, as Tensor -> Union[Tensor, Float] is a backwards compatible change. If you agree we can add the suppress-bc-linter label and I think that should raise them as notifications rather than errors?

Hi @mikaylagawarecki, would you be able to take a look at this comment if you get a chance? I believe adding that label (following the docs here) and rerunning the bc-lint check should fix the last remaining failing check. Thanks! 🙏

@mikaylagawarecki mikaylagawarecki added the suppress-bc-linter Suppresses the failures of API backward-compatibility linter (Lint/bc_linter) label Nov 15, 2024
@mikaylagawarecki
Copy link
Contributor

@pytorchbot merge -r

@pytorch-bot pytorch-bot bot added the ciflow/trunk Trigger trunk jobs on your pull request label Nov 15, 2024
@pytorchmergebot
Copy link
Collaborator

@pytorchbot started a rebase job onto refs/remotes/origin/viable/strict. Check the current status here

@pytorchmergebot
Copy link
Collaborator

Successfully rebased gaussian-nll-loss-scalar-var onto refs/remotes/origin/viable/strict, please pull locally before adding more changes (for example, via git checkout gaussian-nll-loss-scalar-var && git pull --rebase)

@pytorchmergebot pytorchmergebot force-pushed the gaussian-nll-loss-scalar-var branch from 7c7cc82 to 6da0ea9 Compare November 15, 2024 23:10
@pytorchmergebot
Copy link
Collaborator

Merge started

Your 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

Advanced Debugging
Check the merge workflow status
here

@pytorchmergebot
Copy link
Collaborator

Merge failed

Reason: 3 mandatory check(s) failed. The first few are:

Dig deeper by viewing the failures on hud

Details for Dev Infra team Raised by workflow job

Failing merge rule: Core Maintainers

@michael-diggin
Copy link
Contributor Author

michael-diggin commented Nov 20, 2024 •
edited
Loading

Hi @mikaylagawarecki - sorry to bother you again (and hopefully last time!).
The one failing check looks like an unrelated flake based on the pytorch-bot comment here. Is it safe to ignore it, or can we rerun the failed test?

@mikaylagawarecki
Copy link
Contributor

@pytorchbot merge

@pytorchmergebot
Copy link
Collaborator

Merge started

Your 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

Advanced Debugging
Check the merge workflow status
here

@michael-diggin michael-diggin deleted the gaussian-nll-loss-scalar-var branch December 4, 2024 07:52
pobin6 pushed a commit to pobin6/pytorch that referenced this pull request Dec 5, 2024
@michael-diggin
Gaussian nll loss scalar variance support (pytorch#138931)
4157602 
Fixes pytorch#138747

Adds support for `variance` being a Tensor or a float in `gaussian_nll_loss` to avoid a cpu-gpu sync point in the loss function, when the variance is a static tensor like `<scalar>*torch.ones_like(input)`

Pull Request resolved: pytorch#138931
Approved by: https://github.com/mikaylagawarecki
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@mikaylagawarecki mikaylagawarecki mikaylagawarecki approved these changes

@albanD albanD Awaiting requested review from albanD albanD is a code owner

@jbschlosser jbschlosser Awaiting requested review from jbschlosser jbschlosser is a code owner

Assignees

No one assigned

Labels

ciflow/trunk Trigger trunk jobs on your pull request Merged open source release notes: nn release notes category suppress-bc-linter Suppresses the failures of API backward-compatibility linter (Lint/bc_linter) topic: not user facing topic category triaged This issue has been looked at a team member, and triaged and prioritized into an appropriate module

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

Speed up gaussian_nll_loss when variance is always the same scalar value

5 participants

@michael-diggin @mikaylagawarecki @pytorchmergebot @soulitzer @pytorchbot