[Klaud Cold] Update kimik2.5-int4-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: kimik2.5-int4-mi325x-vllm

Test plan

• full-sweep-enabled sweep passes.

🤖 Generated with Claude Code

---

Note

Medium Risk
Single-config container bump can shift MI325X benchmark throughput or stability until the planned full sweep passes; no serving-flag changes in this diff.

Overview
Bumps the Kimi K2.5 INT4 MI325X vLLM benchmark container from vllm/vllm-openai-rocm:v0.21.0 to v0.22.0 in .github/configs/amd-master.yaml. Model (moonshotai/Kimi-K2.5), TP8 fixed-seq-len sweeps, and other recipe fields are unchanged.

Adds a matching perf-changelog.yaml entry for kimik2.5-int4-mi325x-vllm so the image bump is tracked with the PR link (placeholder until merge).

^{Reviewed by Cursor Bugbot for commit 8d377c6. 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=26696246463
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=26696246463

[claude]

LGTM, straightforward vLLM ROCm image bump (v0.21.0 → v0.22.0) for a single recipe with matching perf-changelog entry.

Extended reasoning...

Overview

The PR bumps the vLLM ROCm image for the kimik2.5-int4-mi325x-vllm recipe from v0.21.0 to v0.22.0 in .github/configs/amd-master.yaml, and adds a corresponding entry to perf-changelog.yaml. No other recipes are touched, and no logic/code paths change.

Security risks

None. The change is a YAML version string update to a public vLLM ROCm container tag plus a documentation/changelog append. No auth, crypto, secrets, network endpoints, or permission boundaries are affected.

Level of scrutiny

Low. This is a mechanical, single-recipe image tag bump — exactly the pattern this repo handles regularly via the Klaud Cold automation, and the full-sweep-enabled label ensures the sweep validates runtime correctness. The scope matches the PR title/description.

Other factors

The bug hunting system found no issues. The perf-changelog entry is well-formed and links back to PR #1614. The sibling kimik2.5-int4-mi355x-vllm and kimik2.5-int4-mi300x-vllm entries remain on v0.21.0, which appears intentional given the recipe-by-recipe bump cadence visible elsewhere in the file.

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=26696247974
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=26696247974