[AMD] Update DSR1 MI355X SGLang image / [AMD] 更新 DSR1 MI355X SGLang 镜像

[Oseltamivir]

No description provided.

[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]

Claude finished @Oseltamivir's task in 1m 15s —— View job

---

Review: LGTM — no blocking issues found

• Read PR diff (configs/amd-master.yaml, perf-changelog.yaml)
• Verify new image is a public Docker Hub reference
• Validate perf-changelog entry placement and config-key correctness
• Post review summary

Single image bump for dsr1-fp4-mi355x-sglang-disagg-8k1k-mtp (lmsysorg/sglang-rocm:v0.5.12.post1-...-20260529 → v0.5.16-rocm720-mi35x-20260729). The perf-changelog.yaml entry is correctly appended at the end of the file, references the matching config key and this PR, and the image follows the public Docker Hub format (sibling configs already use v0.5.16-rocm720-mi35x-20260726).

One note: I couldn't verify the exact tag v0.5.16-rocm720-mi35x-20260729 exists on Docker Hub (network access is restricted in this review environment) — the full-sweep run will confirm it pulls successfully.

[claude]

LGTM — simple image tag bump with a matching changelog entry.

Checked that the new image tag v0.5.16-rocm720-mi35x-20260729 is applied only to dsr1-fp4-mi355x-sglang-disagg-8k1k-mtp (the one entry named in the PR title) and that the perf-changelog entry's config-keys matches that same key with a correct PR link.

Extended reasoning...

Overview

This PR touches two files: configs/amd-master.yaml and perf-changelog.yaml. It bumps the SGLang ROCm image for the single dsr1-fp4-mi355x-sglang-disagg-8k1k-mtp benchmark config entry from v0.5.12.post1-rocm720-mi35x-20260529 to v0.5.16-rocm720-mi35x-20260729, and appends a corresponding perf-changelog entry describing the change (an AITER MLA LDS synchronization fix) with a link back to this PR.

Security risks

None. This is a config-only change (YAML data), no executable code paths, no auth/crypto/permissions logic touched. The image tag references a third-party Docker Hub image, consistent with how all other recipes in this file pin images.

Level of scrutiny

This warrants low scrutiny — it's a mechanical, single-value image tag bump matching the well-established pattern used throughout this file (e.g., the immediately preceding commit did the same for a Kimi-K2.5 image). The changelog entry is properly added with the correct config-key and PR link, matching the file's existing conventions. No search-space, routing, or launcher logic was changed.

Other factors

No prior reviews exist on this PR. The bug-hunting system found no issues, and my own read of the diff confirms the change is scoped exactly as described in the PR title/changelog: only the named config entry's image was updated, with no unintended edits elsewhere in the 5000+ line YAML file.