[Klaud Cold] Update minimaxm2.5-fp4-b300-vllm vLLM image to v0.22.0

[functionstackx]

Summary

Update vLLM image from v0.21.0 to v0.22.0

Recipes touched: minimaxm2.5-fp4-b300-vllm

Test plan

• full-sweep-enabled sweep passes.

🤖 Generated with Claude Code

---

Note

Low Risk
Single-container image pin for an existing benchmark key with no serve flags or infra changes; main risk is benchmark/sweep behavior on the new vLLM release.

Overview
Bumps the MiniMax-M2.5 FP4 B300 vLLM benchmark recipe (minimaxm2.5-fp4-b300-vllm) from vllm/vllm-openai:v0.21.0 to v0.22.0 in nvidia-master.yaml. Model, runner, scenarios, and search space are unchanged; the in-repo note still says B300 reuses the B200 recipe until upstream B300 tuning exists.

Adds a matching perf-changelog.yaml entry for this config key documenting the image update.

^{Reviewed by Cursor Bugbot for commit c6622fe. Bugbot is set up for automated code reviews on this repo. Configure here.}

[github-actions]

Thanks for the contribution! For vLLM & SGLang, please ensure that your recipes is similar to the official vLLM recipes and/or the SGLang cookbook

• https://docs.sglang.io/cookbook/intro
• https://recipes.vllm.ai/

If it is not, please create a PR first before we can merge your single node PR into the master branch. Let's ensure that the documentation is first class such that the entire ML community can benefit from your hard work! Thank you

• https://github.com/vllm-project/recipes
• https://github.com/sgl-project/sglang/tree/main/docs_new

PR authors are responsible for ensuring that after merging, all GitHub Action jobs fully pass. A lot of the time, failures are just flakes and simply re-running the failed jobs will fix it. If re-running failed jobs is attempted, PR authors are responsible for ensuring it passes. See GitHub's docs on re-running failed jobs: https://docs.github.com/en/actions/how-tos/manage-workflow-runs/re-run-workflows-and-jobs#re-running-failed-jobs-in-a-workflow

As a rule of thumb, generally, PR authors should request a review & get a PR approval from the respective companies' CODEOWNERS before requesting a review from core maintainers.

If additional help is needed, PR authors can reach out to core maintainers over Slack.

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=26696242524
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=26696242524

[claude]

LGTM, straightforward vLLM image version bump with matching changelog entry.

Extended reasoning...

Overview

This PR updates the minimaxm2.5-fp4-b300-vllm benchmark recipe in .github/configs/nvidia-master.yaml to use vLLM image v0.22.0 instead of v0.21.0, and adds a corresponding entry in perf-changelog.yaml documenting the bump. No other recipe fields (model, runner, scenarios, serve flags, search space) are modified.

Security risks

None. This is a config-only change to a benchmark recipe; no auth, crypto, permissions, or input-handling code is touched. The new image tag points to an official upstream vLLM release.

Level of scrutiny

Low. This is exactly the class of change suitable for shadow approval: a mechanical, self-contained image pin for a single benchmark key, following the established pattern in this repo (image pin + perf-changelog entry referencing the PR).

Other factors

The PR carries the full-sweep-enabled label, so a full benchmark sweep will validate behavior on the new image before/after merge. The bug hunting system reported no findings. The only other timeline activity is the standard recipe-reminder bot comment.

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=26696243980
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=26696243980