Make the PyTorch walltime benchmarks opt-in via a #bench comment - #4383
Open
kinto0 wants to merge 1 commit into
Open
Make the PyTorch walltime benchmarks opt-in via a #bench comment#4383kinto0 wants to merge 1 commit into
#bench comment#4383kinto0 wants to merge 1 commit into
Conversation
Summary: The `pytorch` walltime benchmarks run on every pull request today. They are the expensive half of the suite — a multi-gigabyte cold start driving the LSP server over all cores. They also fail flame-graph generation often enough that unrelated PRs collect a stream of "Unable to generate the flame graphs" comments from CodSpeed. Make them opt-in: the workflow now also listens for `issue_comment`, and the `walltime` job runs only when a PR gets a `#bench` comment. The `simulation` microbenchmarks are cheap and deterministic, so they keep running on every PR and are excluded from the comment trigger. `issue_comment` fires against the default branch, so the job checks out `refs/pull/<n>/merge` to benchmark the PR rather than main. That makes the `author_association` clause load-bearing rather than defensive: without it, any GitHub user could comment `#bench` on a fork PR and have its code run on our runner with an `id-token: write` token. Differential Revision: D114413106
Contributor
|
@kinto0 has exported this pull request. If you are a Meta employee, you can view the originating Diff in D114413106. |
Contributor
Author
|
#bench |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary:
The
pytorchwalltime benchmarks run on every pull request today. They are the expensive half of the suite — a multi-gigabyte cold start driving the LSP server over all cores. They also fail flame-graph generation often enough that unrelated PRs collect a stream of "Unable to generate the flame graphs" comments from CodSpeed.Make them opt-in: the workflow now also listens for
issue_comment, and thewalltimejob runs only when a PR gets a#benchcomment. Thesimulationmicrobenchmarks are cheap and deterministic, so they keep running on every PR and are excluded from the comment trigger.issue_commentfires against the default branch, so the job checks outrefs/pull/<n>/mergeto benchmark the PR rather than main. That makes theauthor_associationclause load-bearing rather than defensive: without it, any GitHub user could comment#benchon a fork PR and have its code run on our runner with anid-token: writetoken.Differential Revision: D114413106