[Klaud Cold] Update dsv4-fp4-b200-vllm (+mtp) vLLM image to v0.22.0

[functionstackx]

Summary

Update vLLM image to v0.22.0 (mtp from v0.21.0; base already pinned to v0.22.0)

Recipes touched: dsv4-fp4-b200-vllm, dsv4-fp4-b200-vllm-mtp

Test plan

• full-sweep-enabled sweep passes.

🤖 Generated with Claude Code

---

Note

Low Risk
Config-only container image pin for benchmark recipes; no application logic or security-sensitive code changes.

Overview
Bumps the DeepSeek V4 FP4 B200 vLLM MTP benchmark recipe (dsv4-fp4-b200-vllm-mtp) container image from vllm/vllm-openai:v0.21.0 to v0.22.0, aligning the MTP sweep with the non-MTP dsv4-fp4-b200-vllm recipe (already on v0.22.0). No search-space or runner changes.

Adds a perf-changelog.yaml entry for dsv4-fp4-b200-vllm and dsv4-fp4-b200-vllm-mtp noting the image update.

^{Reviewed by Cursor Bugbot for commit 9d152ab. 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]

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=26696208519
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=26696208519

[claude]

LGTM — mechanical image pin bump aligning the mtp recipe with its non-mtp sibling.

Extended reasoning...

Overview

This PR updates the container image for the dsv4-fp4-b200-vllm-mtp benchmark recipe in .github/configs/nvidia-master.yaml from vllm/vllm-openai:v0.21.0 to v0.22.0, matching the version already pinned for the sibling dsv4-fp4-b200-vllm recipe. A corresponding entry is added to perf-changelog.yaml documenting the bump.

Security risks

None. This is a config-only change to benchmark recipe metadata; no application logic, auth/crypto code, or security-sensitive paths are touched. The image reference points to the same upstream vllm/vllm-openai publisher used elsewhere in this file.

Level of scrutiny

Low. This is a single-line image tag bump in benchmark recipe config plus a changelog entry — exactly the kind of mechanical version pin that fits the approval criteria. The full-sweep-enabled label will validate the bump via CI.

Other factors

No bugs flagged by the bug hunter. No outstanding human reviewer comments (only the standard recipe-reminder bot notes). No prior reviews from me on this PR.

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=26696210360
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=26696210360