pytorch / pytorch/executorch

Support auto regression detection and bisect

Open
#8,239 2 comments 0 reactions 3 assignees View on GitHub

@wdvr is already working on this.

Since Feb 20, 2025.

module: benchmark module: user experience triaged
Dominant language
Python
Stars
5k
Forks
1.2k
Avg merge
2d 10h
Merged PRs (30d)
581

Description

🚀 The feature, motivation and pitch

Today if a nightly benchmark jobs fail, it fails silently. Workstream members or oncall won't get notification. Performance regression directions is more tricky, we will have to manually compare the commits from nightly runs to get signal.

Ideally, we would like the infra to send alert automatically on failure or if the perf gap exceed the threshold (10%?). The alert should contain a link to the dashboard with involved commits, models, configs, etc. and job links where stacktrace is recorded. Even better if the infra is capable of bisect the root cause automatically and point to the blamed commit/PR in the alert.

Alternatives

No response

Additional context

No response

RFC (Optional)

No response

cc @huydhn @kirklandsign @shoumikhin @mergennachin @byjlw

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.