Skip to content

feat(local_vllm_model): support vLLM 0.25.1 - #2754

Merged
ananthsub merged 1 commit into
mainfrom
ananthsub/vllm-0251-support
Aug 26, 2026
Merged

feat(local_vllm_model): support vLLM 0.25.1#2754
ananthsub merged 1 commit into
mainfrom
ananthsub/vllm-0251-support

Conversation

@ananthsub

@ananthsub ananthsub commented Aug 25, 2026

Copy link
Copy Markdown
Contributor

Summary

  • upgrade local vLLM and GenRM servers to vLLM 0.25.1 and its matching FlashInfer release
  • isolate new codec-bearing transitive dependencies behind the existing codec installation path
  • derive node-local DP ranks from scheduled placement groups so co-located DP engines use distinct TCPStore port windows

Depends on #2456, which updates Gym for the OpenAI version required by vLLM 0.25.1.

Test plan

  • Run pre-commit hooks on changed files
  • Run focused local vLLM model unit tests
  • Start a DP=2 local vLLM server on two GPUs and complete a chat request
  • Start the 32B GenRM checkpoint with TP=2 and verify model-specific scoring

@copy-pr-bot

copy-pr-bot Bot commented Aug 25, 2026

Copy link
Copy Markdown

Auto-sync is disabled for draft pull requests in this repository. Workflows must be run manually.

Contributors can view more details about this message here.

@ananthsub ananthsub changed the title ananthsub/vllm 0251 support feat(local_vllm_model): support vLLM 0.25.1 Aug 25, 2026
@ananthsub ananthsub linked an issue Aug 25, 2026 that may be closed by this pull request
@ananthsub
ananthsub marked this pull request as ready for review August 25, 2026 07:36
@ananthsub
ananthsub requested a review from a team as a code owner August 25, 2026 07:36
@ananthsub
ananthsub force-pushed the ananthsub/openai-window-bump branch from d0d9f1f to 11cd070 Compare August 25, 2026 16:06
@ananthsub
ananthsub force-pushed the ananthsub/vllm-0251-support branch from 98b372e to 6f13d79 Compare August 25, 2026 16:06
@anwithk
anwithk requested a review from ffrujeri August 25, 2026 16:27
ffrujeri
ffrujeri previously approved these changes Aug 25, 2026
@ananthsub

Copy link
Copy Markdown
Contributor Author

/ok to test c136ef1

@ananthsub
ananthsub force-pushed the ananthsub/openai-window-bump branch from 9097819 to a9fc2b6 Compare August 26, 2026 16:51
@ananthsub
ananthsub force-pushed the ananthsub/vllm-0251-support branch from c136ef1 to 5c5e530 Compare August 26, 2026 16:51
@ananthsub
ananthsub force-pushed the ananthsub/openai-window-bump branch from a9fc2b6 to 548c69a Compare August 26, 2026 17:26
Base automatically changed from ananthsub/openai-window-bump to main August 26, 2026 17:41
Upgrade local and GenRM serving dependencies while preserving co-located data-parallel startup with node-aware ranks.

Signed-off-by: Ananth Subramaniam <ansubramania@nvidia.com>
@ananthsub
ananthsub force-pushed the ananthsub/vllm-0251-support branch from 5c5e530 to 4086752 Compare August 26, 2026 17:44
@ananthsub
ananthsub merged commit 0d9f915 into main Aug 26, 2026
36 checks passed
@ananthsub
ananthsub deleted the ananthsub/vllm-0251-support branch August 26, 2026 20:07
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Bump vllm 0.25.1+

2 participants