[Klaud Cold] Update qwen3.8next-fp8-h100-sglang-agentic-mtp SGLang image to dev-cu13-qwen38-next-local (2026-09-07, digest-pinned)

[functionstackx]

Summary

Move the H100 Qwen3.8-Flash-Next FP8 AgentX MTP recipe from the mutable model-branch tag lmsysorg/sglang:qwen38flashnext to the newest Qwen3.8-Flash-Next model-branch build, lmsysorg/sglang:dev-cu13-qwen38-next-local@sha256:9d2a843c706c74bc259c0d9abf360551eb2734e1e7d255ab012a6965f10480b6, pinned by digest.

• Why not an upstream nightly: Qwen3.8-Flash-Next's architecture (Qwen4ExpForConditionalGeneration) is not in SGLang main, v0.5.19, or today's nightly nightly-dev-cu13-20260907-30705c00. It exists only on SGLang model branches shipped as dev images, so the model-branch lineage is the only one that can serve this recipe.
• Old tag: qwen38flashnext was last re-pushed 2026-09-03T20:54:46Z (digest sha256:5ae58167...), after the recipe merged on 2026-08-27, so nodes with an older squash cache may have been running a different build than nodes that pulled later. The digest pin removes that ambiguity.
• New image: dev-cu13-qwen38-next-local, Docker Hub last pushed 2026-09-07T11:12:42Z, digest sha256:9d2a843c706c74bc259c0d9abf360551eb2734e1e7d255ab012a6965f10480b6. Provenance is a model branch, not upstream main; this sweep is the validation that it still carries the fixes the recipe relies on (NEXTN MTP verify with float32 SSM state on sm_90).
• benchmarks/single_node/agentic/qwen3.8next_fp8_h100_mtp.sh is unchanged: TP8/EP8, mem-fraction 0.75, flashinfer attention, NEXTN MTP at 3 speculative tokens, golden AL 2.32, conc [1, 4, 8, 12, 16].

Recipes touched: qwen3.8next-fp8-h100-sglang-agentic-mtp

Test plan

• full-sweep-enabled sweep passes on cluster:h100-dgxc.

🤖 Generated with Claude Code

---

Note

Low Risk
Image reference and changelog only for one benchmark recipe; no runtime code or auth/data paths.

Overview
Updates qwen3.8next-fp8-h100-sglang-agentic-mtp to use the 2026-09-07 model-branch SGLang image lmsysorg/sglang:dev-cu13-qwen38-next-local, digest-pinned, instead of the mutable qwen38flashnext tag so every node runs the same build.

The change is config-only (plus a perf-changelog.yaml entry for the agentic-coding scenario). Benchmark flags, TP8/EP8 layout, and concurrency grid are unchanged; the noted sweep is what validates the new image still supports the recipe (NEXTN MTP, sm_90 flashinfer, etc.). Upstream SGLang nightlies are not an option for this architecture.

^{Reviewed by Cursor Bugbot for commit e3af7fe. 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

[claude]

🔴 The new entry's pr-link uses the literal string "PRLINK_PLACEHOLDER", but the changelog gate only recognizes "XXX" (or the full .../pull/XXX URL) as a valid placeholder, so validate_perf_changelog.py's validate_added_pr_link will reject this entry on CI and prepare_perf_changelog_merge.py will also raise "unexpected pr-link" when preparing the merge. Fix: use the canonical placeholder "XXX" (or the real PR URL) instead of an invented placeholder string.

Extended reasoning...

utils/validate_perf_changelog.py defines PR_LINK_PLACEHOLDERS = {"XXX", "https://github.com/SemiAnalysisAI/InferenceX/pull/XXX"} and CANONICAL_PR_LINK requiring a numeric PR id. validate_added_pr_link() checks the newly appended entry's pr-link against CANONICAL_PR_LINK when pr_number is None, or against PR_LINK_PLACEHOLDERS/expected link when pr_number is set; "PRLINK_PLACEHOLDER" matches neither, so it raises ChangelogValidationError and the changelog gate CI check fails for this PR. Separately, utils/prepare_perf_changelog_merge.py:109-114 does if link not in PR_LINK_PLACEHOLDERS: raise ValueError(f"appended entry {index+1} has unexpected pr-link {link!r}"), so even if the CI gate were bypassed, the merge-preparation tooling used to rewrite the placeholder to the real PR link at merge time would also fail, blocking the automated pr-link substitution step.

Verification: normal. The appended entry at perf-changelog.yaml:6931 uses pr-link: PRLINK_PLACEHOLDER, an invented string. The only recognized placeholders (utils/validate_perf_changelog.py) are PR_LINK_PLACEHOLDERS = {"XXX", "https://github.com/SemiAnalysisAI/InferenceX/pull/XXX"}, and CONTRIBUTING.md:81 documents .../pull/XXX. At merge time utils/merge_with_reuse.sh:187 runs… | normal (with a…

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=34174934041
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=34174934041