[Klaud Cold] Update gptoss-fp4-mi325x-vllm vLLM ROCm image to v0.22.0

[functionstackx]

Summary

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

Recipes touched: gptoss-fp4-mi325x-vllm

Test plan

• full-sweep-enabled sweep passes.

🤖 Generated with Claude Code

---

Note

Low Risk
Config-only container tag bump for one benchmark recipe; no auth, serving logic, or sweep matrix changes.

Overview
Bumps the GPT-OSS 120B FP4 MI325X vLLM benchmark container from vllm/vllm-openai-rocm:v0.21.0 to v0.22.0 for config key gptoss-fp4-mi325x-vllm in .github/configs/amd-master.yaml. Model, runner, and fixed-seq-len search space are unchanged.

Adds a matching perf-changelog.yaml entry documenting the ROCm image update and linking PR #1622.

^{Reviewed by Cursor Bugbot for commit 01fdcd0. 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=26696262993
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=26696262993

[claude]

LGTM, simple image version bump for a single recipe with matching changelog entry.

Extended reasoning...

Overview

This PR bumps the vLLM ROCm image for the gptoss-fp4-mi325x-vllm recipe in .github/configs/amd-master.yaml from v0.21.0 to v0.22.0, and adds a corresponding entry to perf-changelog.yaml. The diff is two lines of substantive change.

Security risks

None. This is a benchmark/sweep configuration change pointing to a versioned image from the same upstream registry (vllm/vllm-openai-rocm) already used throughout this config.

Level of scrutiny

Low. Image version bumps in the AMD master config are mechanical, well-precedented, and gated by the full-sweep-enabled sweep that runs as part of PR validation. Other recipes in the same file already use v0.21.0 and v0.22.0 patterns. The bug-hunting system found no issues.

Other factors

Single-recipe scope, accompanying changelog entry follows the established format, and PR is labeled full-sweep-enabled so the runtime behavior will be validated automatically.

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=26696264378
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=26696264378

[functionstackx]

/reuse-sweep-run

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=26698276506
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=26698276506