Repository navigation
[Dev] Remove mandatory scheduler reservations for GPU validation - #173
Merged
Merged
Conversation
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Purpose
Remove the mandatory scheduler-reservation prerequisite for GPU validation. An explicit GPU-testing request can proceed on an available, explicitly selected device even when no scheduler is installed.
Update the review skill, pytest help/error/skip wording, marker description, and current serving/rerun instructions. Keep
--run-gpu, explicitCUDA_VISIBLE_DEVICES, authorized experiment scope, and run budgets. Historical benchmark descriptions remain unchanged.Required human contributor answers
1. What outcome are you trying to achieve?
Allow requested GPU validation without blocking on a scheduler reservation.
2. What is the highest-leverage variable for that outcome?
The mandatory reservation prerequisite in the review workflow.
3. How will this work move that variable rather than create busywork?
Remove the prerequisite and align the associated instructions and pytest wording with explicit device selection.
4. Who is the customer or user, and what do they actually need?
Contributors and reviewers running requested GPU tests on machines without a configured scheduler.
PR necessity and bugfix evidence
A PR is needed to remove the repository-wide requirement and its contradictory usage instructions. The existing review rule forbids unreserved execution, while the pytest guard only requires an explicit device selection. This is a workflow-policy correction; inference behavior and test gating logic are unchanged.
Test Plan
Run
pre-commit run --all-filesandgit diff --check. Inspect remaining reservation references to distinguish current requirements from historical experiment facts. Runtime performance analysis is not applicable to policy/documentation and message-only changes.vllm-rlt Version / Commit: base
444638beca89e95af6efb34a39fd3edf01c4020bTest Result
pre-commit run --all-files: passed Ruff lint, Ruff format, and the CPU regression suite.git diff --check: passed.