[CI] Fix slow grpo CI #3693
Merged
[CI] Fix slow grpo CI #3693
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
|
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
|
@ArthurZucker the paged attention CI was failing since it didnt find flash_attn, so i replaced the check with |
There was a problem hiding this comment.
Pull Request Overview
This PR enhances CI for the GRPO trainer by preferring FlashAttention v2 when available, updates a comment for consistency in the vLLM serve script, and fixes a markdown warning in the SFT trainer docs.
- Import and use
is_flash_attn_2_availableinstead of solely relying ontorch.cuda.is_available - Update a misaligned example comment in
vllm_serve.pyto match actual dtype usage - Wrap the
{% generation %}tag in a raw block in the SFT trainer docs warning
Reviewed Changes
Copilot reviewed 3 out of 3 changed files in this pull request and generated 1 comment.
| File | Description |
|---|---|
| trl/trainer/grpo_trainer.py | Added import of is_flash_attn_2_available and replaced the CUDA check for attention |
| trl/scripts/vllm_serve.py | Corrected comment to show "torch.float32" as a string in collective_rpc example |
| docs/source/sft_trainer.md | Changed warning block to use a raw tag around {% generation %} |
shirinyamani
approved these changes
Jul 4, 2025
|
Yes completely! Sorry |
marcandrelarochelle
pushed a commit
to marcandrelarochelle/trl
that referenced
this pull request
Jul 29, 2025
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Outdated suggestions cannot be applied.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.
What does this PR do?
We now check if flash_attn is installed and if not use the
sdpa_pagedin the transformers paged attention slow testBefore submitting
Pull Request section?
to it if that's the case.
Who can review?
Anyone in the community is free to review the PR once the tests have passed. Feel free to tag
members/contributors who may be interested in your PR.