[AMD][MI35X] Bump glm5.2-fp4-mi355x-sglang-agentic-mtp image to v0.5.19-rocm10-20260908 - #2917
[AMD][MI35X] Bump glm5.2-fp4-mi355x-sglang-agentic-mtp image to v0.5.19-rocm10-20260908#2917jiaryang wants to merge 3 commits into
Conversation
…19-rocm10-20260908 Bump image from lmsysorg/sglang-rocm:v0.5.16-rocm720-mi35x-20260728 to lmsysorg/sglang-rocm:v0.5.19-rocm10-mi35x-20260908. Recipe unchanged. Co-authored-by: Cursor <cursoragent@cursor.com>
|
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 For PR verification, add the 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 如需进行 PR 验证，请为此 PR 添加 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 |
Co-authored-by: Cursor <cursoragent@cursor.com>
![cursor[bot]](https://avatars.githubusercontent.com/in/1210556?s=60&v=4){kind=small}
There was a problem hiding this comment.
Cursor Bugbot has reviewed your changes using default effort and found 1 potential issue.
❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.
Want higher recall? High effort reviews run extra passes and find more bugs. A team admin can switch effort levels in the Cursor dashboard.
Reviewed by Cursor Bugbot for commit 3b5875a. Configure here.
| # SA selects the Pareto-optimal arm per concurrency point. | ||
| glm5.2-fp4-mi355x-sglang-agentic-mtp: | ||
| image: lmsysorg/sglang-rocm:v0.5.16-rocm720-mi35x-20260728 | ||
| image: lmsysorg/sglang-rocm:v0.5.19-rocm10-mi35x-20260908 |
There was a problem hiding this comment.
Pinned image tag does not exist
High Severity
The new pin lmsysorg/sglang-rocm:v0.5.19-rocm10-mi35x-20260908 is not a published Docker Hub tag. Published v0.5.19 MI35X images use the 20260905 suffix, including v0.5.19-rocm10-mi35x-20260905. Jobs will fail at image pull before any serving starts.
Additional Locations (1)
Reviewed by Cursor Bugbot for commit 3b5875a. Configure here.
![claude[bot]](https://avatars.githubusercontent.com/in/1236702?s=60&v=4){kind=small}
There was a problem hiding this comment.
I reviewed this PR and didn't find any bugs. Because it's a CODEOWNER-gated config file and the new tag introduces an unprecedented rocm10 component with the full sweep still unchecked, a human look would still be worthwhile before merge.
What was reviewed:
- Confirmed the image bump in
configs/amd-master.yamlmatches the tag names recorded in the newperf-changelog.yamlentry, and that the entry is appended only at the file's tail with no historical bytes touched. - Cross-checked the new tag's
rocm10component against the other 18sglang-rocm/sglangimage references inamd-master.yaml— none use it (all userocm700/rocm720/rocm7.2.x), matching the "if zero recipes use it, be suspicious" heuristic inKLAUD_DEBUG.md§6; this sandbox has no network egress to Docker Hub, so the tag's existence could not be independently confirmed. - Noted
pr-link: XXXis a known placeholder per the PR's own test-plan checkbox, to be filled in before merge/reuse.
Extended reasoning...
Overview
The diff is exactly as described: a single-line image bump in configs/amd-master.yaml for the glm5.2-fp4-mi355x-sglang-agentic-mtp recipe (v0.5.16-rocm720-mi35x-20260728 → v0.5.19-rocm10-mi35x-20260908), plus a matching append-only entry in perf-changelog.yaml. No other lines are touched in either file.
Security risks
None identified. This is a Docker image tag string change in a YAML config plus a changelog append; there is no code path, auth, or data-handling logic involved.
Level of scrutiny
configs/amd-master.yaml is CODEOWNER-gated (.github/CODEOWNERS lists five named owners for this exact file), which per the approval guidelines means this should not be auto-approved regardless of how small the diff looks. Separately, the new tag's rocm10 component is unique among all 19 sglang/sglang-rocm image references in this file (everything else uses rocm700/rocm720/rocm7.2.x), which is precisely the "be suspicious" signal KLAUD_DEBUG.md §6 calls out for invented/incorrect image tags. I could not reach Docker Hub from this sandbox to confirm the tag exists, and the PR's own test plan shows the full AgentX sweep as not yet run. That combination — CODEOWNER path plus an unverified, pattern-breaking tag component plus a pending sweep — is enough that a human (ideally a CODEOWNER, who can also verify the tag pulls) should look before merge, even though the bug-hunting pass itself found no concrete bug.
Other factors
The pr-link: XXX placeholder is a routine, expected leftover (the PR description itself flags it as a pre-merge TODO), so it is not a functional defect on its own, just noted for completeness.
This review covers commit 5f9b992, which is no longer the latest commit on this pull request; later commits are not covered by it.
…5x-sglang-agentic-mtp-20260908 Co-authored-by: Cursor <cursoragent@cursor.com> # Conflicts: # perf-changelog.yaml
|
see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=34340229025 |
|
see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=34367675155 |
1 participant
Summary
glm5.2-fp4-mi355x-sglang-agentic-mtplmsysorg/sglang-rocm:v0.5.16-rocm720-mi35x-20260728→lmsysorg/sglang-rocm:v0.5.19-rocm10-mi35x-20260908Recipe unchanged.
Unofficial local vs CI MI355X TP4/EP4. Δ = 0908/CI (higher is better).
Test plan
validate_perf_changelog.pyagainstupstream/mainfull-sweep-fail-fast)