[Klaud Cold] Update gptoss-fp4-mi300x-vllm vLLM ROCm image to v0.22.0

[functionstackx]

Summary

Update vLLM ROCm image from v0.17.0 to v0.22.0

Recipes touched: gptoss-fp4-mi300x-vllm

Test plan

• full-sweep-enabled sweep passes.

🤖 Generated with Claude Code

---

Note

Medium Risk
Large vLLM ROCm version jump on a live benchmark recipe can change throughput, stability, or ROCm-specific behavior until the full-sweep validation passes.

Overview
Bumps the gptoss-fp4-mi300x-vllm benchmark recipe to vllm/vllm-openai-rocm:v0.22.0 (from v0.17.0) in .github/configs/amd-master.yaml. Model, runner, TP/concurrency search space, and scenarios are unchanged.

Adds a matching perf-changelog.yaml entry for config key gptoss-fp4-mi300x-vllm so the image bump is tracked in perf history.

^{Reviewed by Cursor Bugbot for commit ca62d9c. 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=26696261152
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=26696261152

[claude]

LGTM — single-recipe vLLM ROCm image bump with matching perf-changelog entry.

Extended reasoning...

Overview

This PR bumps the vLLM ROCm image for the gptoss-fp4-mi300x-vllm recipe in .github/configs/amd-master.yaml from v0.17.0 to v0.22.0, and appends a corresponding entry to perf-changelog.yaml. Scope is two lines of YAML plus a 5-line changelog block; no other recipes or code paths touched.

Security risks

None. The change is a container image tag bump in a benchmark sweep config, not in production-serving code. The target image vllm/vllm-openai-rocm:v0.22.0 follows the same registry/tag pattern used elsewhere in the file.

Level of scrutiny

Low. This is the standard mechanical image-bump pattern routinely applied across recipes in this config, and the full-sweep-enabled label means the sweep CI will validate the bump before any downstream consumer sees results.

Other factors

No bugs were flagged by the bug hunting system, the diff is self-contained and reversible, and the perf-changelog entry correctly references the PR. Nothing here warrants human review.

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=26696262596
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=26696262596