[Klaud Cold] Update dsv4-fp4-mi355x-vllm-agentic-mtp vLLM ROCm image to nightly-e962733e08d10f7ca65dac4df99e116460b8b174

[functionstackx]

Summary

Update vLLM ROCm image for dsv4-fp4-mi355x-vllm-agentic-mtp from nightly-311b3513af33bc29b4acb2fde2e9313e5e9966a0 (2026-08-17) to nightly-e962733e08d10f7ca65dac4df99e116460b8b174 (2026-09-05, digest sha256:511755968434e26da590c3398f1e8a6399be4416226d5779ce92fb0655811a9c).

Recipes touched: dsv4-fp4-mi355x-vllm-agentic-mtp

Test plan

• full-sweep-enabled sweep passes.

🤖 Generated with Claude Code

---

Note

Low Risk
Config-only image pin with changelog documentation; no application code or sweep topology changes.

Overview
Repins the vLLM ROCm serving image for dsv4-fp4-mi355x-vllm-agentic-mtp from nightly-311b3513… (2026-08-17) to nightly-e962733e… (2026-09-05). Model, runner, agentic-coding scenario, and TP8 search space are unchanged—only the container tag in configs/amd-master.yaml.

Adds a matching perf-changelog.yaml entry for config key dsv4-fp4-mi355x-vllm-agentic-mtp / agentic-coding, noting the new digest and upstream vLLM commit.

^{Reviewed by Cursor Bugbot for commit b97034c. Bugbot is set up for automated code reviews on this repo. Configure here.}

[github-actions]

Thanks for the contribution! Please reach out to respective companies' CODEOWNER to fill in the latest PR_REVIEW_CHECKLIST.md before pinging core maintainer on Slack for review. In order for the signoff PR check bot to trigger, you must follow the PR_REVIEW_CHECKLIST.md template correctly, including the phrase As a PR reviewer and CODEOWNER, I have reviewed this and have.

For PR verification, add the full-sweep-fail-fast label (strongly recommended) to this PR — the benchmark sweep only runs on labeled PRs. Use full-sweep-enabled only if you need matrix jobs to keep running past a failure.

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. See GitHub's docs on re-running failed jobs

---

感谢你的贡献！请联系相应公司的 CODEOWNER 填写最新的 PR_REVIEW_CHECKLIST.md，然后再在 Slack 上联系核心维护者进行审阅。为了触发 signoff PR 检查机器人，你必须正确遵循 PR_REVIEW_CHECKLIST.md 模板，包括保留英文语句 As a PR reviewer and CODEOWNER, I have reviewed this and have。

如需进行 PR 验证，请为此 PR 添加 full-sweep-fail-fast 标签（强烈推荐）— 基准测试 sweep 仅在带有标签的 PR 上运行。仅当需要矩阵任务在失败后继续运行时才使用 full-sweep-enabled。

PR 作者有责任确保合并后所有 GitHub Action 任务完全通过。 很多时候失败只是偶发抖动（flake），重新运行失败的任务即可解决。参见 GitHub 关于重新运行失败任务的文档

[github-actions]

Thanks for the contribution! Please reach out to respective companies' CODEOWNER to fill in the latest PR_REVIEW_CHECKLIST.md before pinging core maintainer on Slack for review. In order for the signoff PR check bot to trigger, you must follow the PR_REVIEW_CHECKLIST.md template correctly, including the phrase As a PR reviewer and CODEOWNER, I have reviewed this and have.

For PR verification, add the full-sweep-fail-fast label (strongly recommended) to this PR — the benchmark sweep only runs on labeled PRs. Use full-sweep-enabled only if you need matrix jobs to keep running past a failure.

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. See GitHub's docs on re-running failed jobs

---

感谢你的贡献！请联系相应公司的 CODEOWNER 填写最新的 PR_REVIEW_CHECKLIST.md，然后再在 Slack 上联系核心维护者进行审阅。为了触发 signoff PR 检查机器人，你必须正确遵循 PR_REVIEW_CHECKLIST.md 模板，包括保留英文语句 As a PR reviewer and CODEOWNER, I have reviewed this and have。

如需进行 PR 验证，请为此 PR 添加 full-sweep-fail-fast 标签（强烈推荐）— 基准测试 sweep 仅在带有标签的 PR 上运行。仅当需要矩阵任务在失败后继续运行时才使用 full-sweep-enabled。

PR 作者有责任确保合并后所有 GitHub Action 任务完全通过。 很多时候失败只是偶发抖动（flake），重新运行失败的任务即可解决。参见 GitHub 关于重新运行失败任务的文档

[github-actions]

Thanks for the contribution! Please reach out to respective companies' CODEOWNER to fill in the latest PR_REVIEW_CHECKLIST.md before pinging core maintainer on Slack for review. In order for the signoff PR check bot to trigger, you must follow the PR_REVIEW_CHECKLIST.md template correctly, including the phrase As a PR reviewer and CODEOWNER, I have reviewed this and have.

For PR verification, add the full-sweep-fail-fast label (strongly recommended) to this PR — the benchmark sweep only runs on labeled PRs. Use full-sweep-enabled only if you need matrix jobs to keep running past a failure.

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. See GitHub's docs on re-running failed jobs

---

感谢你的贡献！请联系相应公司的 CODEOWNER 填写最新的 PR_REVIEW_CHECKLIST.md，然后再在 Slack 上联系核心维护者进行审阅。为了触发 signoff PR 检查机器人，你必须正确遵循 PR_REVIEW_CHECKLIST.md 模板，包括保留英文语句 As a PR reviewer and CODEOWNER, I have reviewed this and have。

如需进行 PR 验证，请为此 PR 添加 full-sweep-fail-fast 标签（强烈推荐）— 基准测试 sweep 仅在带有标签的 PR 上运行。仅当需要矩阵任务在失败后继续运行时才使用 full-sweep-enabled。

PR 作者有责任确保合并后所有 GitHub Action 任务完全通过。 很多时候失败只是偶发抖动（flake），重新运行失败的任务即可解决。参见 GitHub 关于重新运行失败任务的文档

[claude]

Beyond the ruled-out placeholder pr-link (expected mid-flow per the repo's own nuke.md automation, which patches it in via amend+force-push once the PR URL is known), I checked the changelog entry format and image tag against neighboring entries — both match existing conventions, and no other issues stood out.

Extended reasoning...

This is a pure config/changelog diff: a single Docker image tag repin in configs/amd-master.yaml and one appended, well-formed entry in perf-changelog.yaml. No source code, auth, or security-sensitive paths are touched, so there are no meaningful security risks to flag.

The only candidate concern — the pr-link field being the literal "PRLINK_PLACEHOLDER" — was already investigated and ruled out by the bug hunter's own findings list, and I independently confirmed why: grepping the repo shows .claude/commands/nuke.md is the tool that generates changelog entries with exactly this placeholder text, later replaced by a real PR URL via git commit --amend + force-push once the PR is created. This is a known, self-healing step in the repo's own workflow rather than a defect introduced by this change.

Since a finding-adjacent concern was examined and ruled out with a concrete reason beyond what's already visible, I'm using the defer option's narrow exception to record that, per the instructions, rather than staying fully silent or approving outright.

This review covers commit ad5a405, which is no longer the latest commit on this pull request; later commits are not covered by it.

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=33986337598
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=33986337598

[functionstackx]

/reuse-sweep-run 33986337598

[functionstackx]

/reuse-sweep-run