Skip to content

Conversation

@hl475
Copy link
Contributor

@hl475 hl475 commented Oct 30, 2025

Purpose

Corresponding PR on vllm-project/ci-infra - vllm-project/ci-infra#200

This PR adds autorun_on_main: true into test label: LM Eval Small Models label: LM Eval Large Models (H100). As explained from PR on ci-infra side, we want to maintain limited/heavy tests to guard the trunk health at some degree.

The goal of this PR is to add a new feature - if a test/label has autorun_on_main as True AND the PR has been merged into main, then the corresponding test/label will run automatically (so no need to manually click to unblock).

The hope of the feature is adding some selected tests/labels to run automatically on PR to maintain trunk health. Some alternatives could be (1) merge queue, (2) new symbol other than ready. I am open to alternative, but it could change user experience a bit.

For any tests on broad to autorun_on_main, it will likely increase the CI cost. Please keep this in mind.

Test Plan

CI

Test Result

https://buildkite.com/vllm/ci/builds/38581/steps/canvas?sid=019a7696-a611-4e2f-8ce6-d0b873f4f5fa


Essential Elements of an Effective PR Description Checklist
  • The purpose of the PR, such as "Fix some issue (link existing issues this PR will resolve)".
  • The test plan, such as providing test command.
  • The test results, such as pasting the results comparison before and after, or e2e results
  • (Optional) The necessary documentation update, such as updating supported_models.md and examples for a new model.
  • (Optional) Release notes update. If your change is user facing, please update the release notes draft in the Google Doc.

@mergify mergify bot added the ci/build label Oct 30, 2025
@hl475 hl475 marked this pull request as ready for review October 31, 2025 04:10
@hl475 hl475 marked this pull request as draft October 31, 2025 04:12
@hl475 hl475 force-pushed the enable_autorun_on_ready_for_eval branch 2 times, most recently from c787a96 to ea7af9f Compare October 31, 2025 16:46
@hl475 hl475 changed the title [WIP] enable_autorun_on_ready_for_eval [CI] enable_autorun_on_ready_for_eval Nov 1, 2025
@hl475 hl475 changed the title [CI] enable_autorun_on_ready_for_eval [CI] Introduce ci_always_run feature Nov 1, 2025
@hl475 hl475 marked this pull request as ready for review November 1, 2025 00:10
@hl475
Copy link
Contributor Author

hl475 commented Nov 1, 2025

cc @yeqcharlotte @khluu

@hl475 hl475 marked this pull request as draft November 5, 2025 07:56
@hl475 hl475 force-pushed the enable_autorun_on_ready_for_eval branch from ea7af9f to c5c9d34 Compare November 7, 2025 17:54
@hl475 hl475 changed the title [CI] Introduce ci_always_run feature [CI] Introduce autorun_on_main feature Nov 7, 2025
@hl475 hl475 marked this pull request as ready for review November 7, 2025 17:57
@khluu
Copy link
Collaborator

khluu commented Nov 11, 2025

I'm not sure if the H100 queue capacity is enough to run per commit on main...

@hl475 hl475 force-pushed the enable_autorun_on_ready_for_eval branch from 6d5e0a5 to d1c9954 Compare November 12, 2025 05:38
@hl475
Copy link
Contributor Author

hl475 commented Nov 12, 2025

add autorun_on_main: true into label: LM Eval Small Models per suggestion

Copy link
Collaborator

@yeqcharlotte yeqcharlotte left a comment

Choose a reason for hiding this comment

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

Let's see how it goes!

@yeqcharlotte yeqcharlotte enabled auto-merge (squash) November 12, 2025 06:57
@github-actions github-actions bot added the ready ONLY add when PR is ready to merge/full CI is needed label Nov 12, 2025
@yeqcharlotte yeqcharlotte merged commit c748355 into vllm-project:main Nov 12, 2025
33 checks passed
@hl475 hl475 deleted the enable_autorun_on_ready_for_eval branch November 12, 2025 09:09
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

ci/build ready ONLY add when PR is ready to merge/full CI is needed

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants