[Klaud Cold] Update kimik2.5-fp4-mi355x-vllm vLLM ROCm image to v0.22.0

[functionstackx]

Summary

Update vLLM ROCm image from v0.21.0 to v0.22.0

Recipes touched: kimik2.5-fp4-mi355x-vllm

Test plan

• full-sweep-enabled sweep passes.

🤖 Generated with Claude Code

---

Note

Low Risk
Config-only container tag bump for a single benchmark recipe; no application code or sweep matrix changes.

Overview
Bumps the MI355X fixed-seq-len benchmark recipe kimik2.5-fp4-mi355x-vllm from vllm/vllm-openai-rocm:v0.21.0 to v0.22.0. Model (amd/Kimi-K2.5-MXFP4), runner, and TP/concurrency search space are unchanged.

Adds a perf-changelog.yaml entry documenting the image bump and linking PR #1616.

^{Reviewed by Cursor Bugbot for commit 19c46ee. 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=26696251544
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=26696251544

[claude]

LGTM — straightforward image tag bump (v0.21.0 → v0.22.0) for one recipe with matching changelog entry.

Extended reasoning...

Overview

This PR bumps the vLLM ROCm image for the kimik2.5-fp4-mi355x-vllm recipe from v0.21.0 to v0.22.0 in .github/configs/amd-master.yaml, and adds the corresponding entry to perf-changelog.yaml. The change is two lines of YAML plus a six-line changelog stanza, scoped to exactly one config key.

Security risks

None. This is a benchmark-recipe image tag bump pointing to a public upstream vLLM container; it does not touch auth, crypto, permissions, secrets, or any production-serving path.

Level of scrutiny

Minimal. This is the canonical mechanical config change pattern in this repo — a single recipe's image tag is bumped to a newer upstream tagged release, the changelog reflects exactly that, and the full-sweep-enabled label means CI will run the actual recipe sweep as the functional gate.

Other factors

The bug hunting system found nothing, there are no outstanding reviewer comments to address, only the routine recipe-reminder bot comment is on the timeline, and CI sweep coverage is the right validator for an image-tag bump like this.

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=26696253037
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=26696253037

[functionstackx]

/reuse-sweep-run

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=26701267138
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=26701267138