[CI] Disable CUDA graphs for GLM PCP evals - #54745
Merged
Merged
Conversation
MRV2 sparse MLA PCP does not support CUDA graphs. Run both GLM-5.2 PCP accuracy configurations with cudagraph_mode NONE so the B200 LM Eval PCP lane can start and execute GSM8K.\n\nCo-authored-by: Codex <noreply@openai.com> Signed-off-by: Kevin Luu <51931015+khluu@users.noreply.github.com>
khluu
marked this pull request as ready for review
September 1, 2026 09:52
khluu
requested review from
AndreasKaratzas,
mgoin and
vadiklyutiy
as code owners
September 1, 2026 09:52
![claude[bot]](https://avatars.githubusercontent.com/in/1236702?s=60&v=4){kind=small}
njhill
approved these changes
Sep 1, 2026
njhill
enabled auto-merge (squash)
September 1, 2026 19:09
Member
Author
|
/ci run |
|
✅ Triggered Buildkite CI #86694 for commit |
Contributor
|
Thanks for fixing this! |
khluu
added a commit
that referenced
this pull request
Sep 2, 2026
Signed-off-by: Kevin Luu <51931015+khluu@users.noreply.github.com>
mylibrar
pushed a commit
to tanyuqian/vllm
that referenced
this pull request
Sep 3, 2026
Signed-off-by: Kevin Luu <51931015+khluu@users.noreply.github.com>
sheralskumar
pushed a commit
to sheralskumar/vllm
that referenced
this pull request
Sep 8, 2026
Signed-off-by: Kevin Luu <51931015+khluu@users.noreply.github.com>
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
3 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
cudagraph_modetoNONEfor both GLM-5.2 NVFP4 PCP eval configurationsWhy
MRV2 sparse MLA PCP does not support CUDA graphs yet. Nightly build #86511 failed both selected GLM PCP configurations during engine initialization with:
This changes only those two test configurations. PR #53896 is not the source of this failure; it added the separate Qwen3.8 accuracy lane. The failing PCP configs were changed to
PIECEWISEby PR #53869.Duplicate check
Searched open PRs for
Qwen3.8-Flash-Next PCP,cudagraph_mode=NONE PCP,GLM-5.2 PCP, andLM Eval PCP. No open PR addresses this configuration fix.Validation
git diff --checkserver_argswithshlex, decoded the compilation config as JSON, and assertedcudagraph_mode == "NONE"Model evaluation was not run locally because this lane requires four B200 GPUs and the GLM-5.2-NVFP4 model. The exact
lm-eval-pcp-4xb200Buildkite lane is the required validation.AI assistance was used to investigate the CI failure and prepare this change. A human submitter must review every changed line and the validation evidence before marking the PR ready.