Skip to content

Conversation

@izaitsevfb
Copy link
Contributor

Currently, we have a slight overlap between job and test Signals. Specifically, when there is any non-test related failure in the job within the lookback window (e.g. any infra flake), the job signal is extracted, and it extracts ANY job failures, including test failures as the job signal. This potentially leads to the duplication, when the same test failure is processed both as test and job-track Signals, and it increases the chance of false positives.

This PR makes the separation explicit:

  1. test failures are extracted as test-track Signals
  2. job-track Signals deliberately ignore job failures caused exclusively by tests (such events are extracted as successes)

The intended outcome of this change:

  1. job-track signals will only try to revert infra-breaking changes and ignore test failures
  2. test-track signal processing remains as it is

@pytorch-bot pytorch-bot bot added the ci-no-td label Oct 22, 2025
@meta-cla meta-cla bot added the CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. label Oct 22, 2025
@vercel
Copy link

vercel bot commented Oct 22, 2025

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment
Project Deployment Preview Updated (UTC)
torchci Ignored Ignored Oct 22, 2025 7:46pm

@izaitsevfb izaitsevfb merged commit 6498804 into main Oct 22, 2025
6 checks passed
@izaitsevfb izaitsevfb deleted the autorevert-tighten-job-track-signals branch October 22, 2025 23:24
jeanschmidt added a commit that referenced this pull request Nov 10, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

ci-no-td CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants