Skip to content

[Klaud Cold] Update qwen3.5-fp8-b200-dynamo-sglang-agentic-disagg-mtp SGLang image to nightly-dev-cu13-20260922-4cbf290f / 将 qwen3.5-fp8-b200-dynamo-sglang-agentic-disagg-mtp 的 SGLang 镜像更新至 nightly-dev-cu13-20260922-4cbf290f - #3375

Closed
Klaud-Cold wants to merge 1 commit into
mainfrom
klaud/auto-e920fad623895413-abb39e882feee984

Conversation

@Klaud-Cold

@Klaud-Cold Klaud-Cold commented Sep 23, 2026 •

Copy link
Copy Markdown
Collaborator

Goal: Update SGLang image from lmsysorg/sglang:nightly-dev-cu13-20260918-20518d85 to lmsysorg/sglang:nightly-dev-cu13-20260922-4cbf290f.
Baseline: 2026-09-19 · lmsysorg/sglang:nightly-dev-cu13-20260918-20518d85
AgentX · PTP4x1/DTP4x1 · Mean latency · Sources: API 1, API 2, API 3, API 4, API 5

Concurrency Total tok/s/GPU ↑ Output tok/s/GPU ↑ TTFT ms ↓ TPOT ms ↓
8 5,531.21 54.96 426.57 2.42
16 12,419.78 106.07 571.48 3.09
24 17,815.92 149.66 731.12 3.19
32 20,703.06 185.19 934.1 3.59
48 27,982.94 283.37 1,401.29 5.18
64 34,825.43 377.64 3,370.29 7.37
Eval Score ↑ Samples
gsm8k/em_strict · c8 97.57% 1,319
gsm8k/em_strict · c16 97.8% 1,319
gsm8k/em_strict · c24 97.8% 1,319
gsm8k/em_strict · c32 97.42% 1,319
gsm8k/em_strict · c48 97.88% 1,319
gsm8k/em_strict · c64 97.95% 1,319
中文

**目标:**将 SGLang 镜像从 lmsysorg/sglang:nightly-dev-cu13-20260918-20518d85 更新为 lmsysorg/sglang:nightly-dev-cu13-20260922-4cbf290f。
**基线:**2026-09-19 · lmsysorg/sglang:nightly-dev-cu13-20260918-20518d85
AgentX · PTP4x1/DTP4x1 · 平均延迟 · 来源: API 1, API 2, API 3, API 4, API 5;数值及异常说明见上表。

…mage to nightly-dev-cu13-20260922-4cbf290f

Bump qwen3.5-fp8-b200-dynamo-sglang-agentic-disagg-mtp from
lmsysorg/sglang:nightly-dev-cu13-20260918-20518d85 to the digest-pinned
lmsysorg/sglang:nightly-dev-cu13-20260922-4cbf290f and keep the six
srt-slurm recipes' model.container equal to the master image.

将 qwen3.5-fp8-b200-dynamo-sglang-agentic-disagg-mtp 的 SGLang 镜像从
lmsysorg/sglang:nightly-dev-cu13-20260918-20518d85 更新为按 digest 固定的
lmsysorg/sglang:nightly-dev-cu13-20260922-4cbf290f,并同步六个 srt-slurm
配方的 model.container 与主配置镜像一致。

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
@github-actions

Copy link
Copy Markdown
Contributor

Thanks for the contribution!

  • Review: If this PR changes files owned by someone other than a repository admin or @SemiAnalysisAI/core, ask one eligible CODEOWNER to complete the latest PR_REVIEW_CHECKLIST.md before contacting a core maintainer on Slack. Follow the template exactly, including As a PR reviewer and CODEOWNER, I have reviewed this and have, so sign-off verification triggers.
  • PR verification: Sweeps only run on labeled PRs. Add full-sweep-fail-fast (strongly recommended); use full-sweep-enabled only when matrix jobs should continue after a failure.
  • After merging: PR authors must ensure all GitHub Actions jobs pass. Transient failures often pass on rerun; see how to rerun failed jobs.
中文

感谢你的贡献!

  • **审阅:**如果 PR 修改的文件归属于仓库管理员及 @SemiAnalysisAI/core 之外的 CODEOWNER,请先联系一位有资格的 CODEOWNER 填写最新的 PR_REVIEW_CHECKLIST.md,再通过 Slack 联系核心维护者。必须严格遵循模板,并保留 As a PR reviewer and CODEOWNER, I have reviewed this and have,才能触发签核验证。
  • **PR 验证:**扫描仅在带有标签的 PR 上运行。强烈建议添加 full-sweep-fail-fast;仅当需要矩阵任务在失败后继续运行时才使用 full-sweep-enabled。
  • **合并后:**PR 作者必须确保所有 GitHub Actions 任务通过。临时性失败通常可以通过重新运行恢复;参见重新运行失败任务的说明。

@Klaud-Cold

Klaud-Cold commented Sep 23, 2026 •

Copy link
Copy Markdown
Collaborator Author

Initial attempt · Failed · Run 35804907428 / attempt 1 · 2026-09-23 01:10 UTC
lmsysorg/sglang:nightly-dev-cu13-20260922-4cbf290f@sha256:504d34fd712ac2656e9a9da363160e49fddb9957b60092893d6b773fc26c0f5f · c8b8d5a30bdc · AgentX · PTP4x1/DTP4x1 · Mean latency
Change: Bump SGLang from commit 20518d85 (2026-09-18) to 4cbf290f (2026-09-22; 295 commits), digest-pinned, with unchanged Dockerfile, CUDA 13.0.3, FlashInfer 0.6.18, sgl-kernel 0.4.7 and Mooncake 0.3.13; all recipe flags still parse, but upstream defaults changed: FlashInfer MoE fused finalize off (#40105), HiCache TMA transfer kernel on (#40278) and the legacy MambaRadixCache class was removed (#40313).

Concurrency Output tok/s/GPU ↑ TTFT ms ↓ TPOT ms ↓
8 N/A N/A N/A
16 N/A N/A N/A
24 N/A N/A N/A
32 N/A N/A N/A
48 N/A N/A N/A
64 N/A N/A N/A

Note: All rows: cancelled.
Note: All rows: request errors unavailable.
Note: All rows: Δ N/A: point unavailable.

Next: Report readiness-blocked and release the family: the shared smoke generator emits eval-all-concs on multi-node agentic eval rows, which the matrix schema forbids, so no updated-image job can start until main fixes it.

中文

初次尝试 · 失败 · Run 35804907428 / attempt 1 · 2026-09-23 01:10 UTC
lmsysorg/sglang:nightly-dev-cu13-20260922-4cbf290f@sha256:504d34fd712ac2656e9a9da363160e49fddb9957b60092893d6b773fc26c0f5f · c8b8d5a30bdc · AgentX · PTP4x1/DTP4x1 · 平均延迟
**变更:**将 SGLang 从提交 20518d85(2026-09-18)更新至 4cbf290f(2026-09-22,共 295 个上游提交),镜像按 digest 固定,Dockerfile、CUDA 13.0.3、FlashInfer 0.6.18、sgl-kernel 0.4.7 与 Mooncake 0.3.13 均未变化;配方全部参数仍可解析,但上游默认值有变:FlashInfer MoE fused finalize 默认关闭(#40105)、HiCache TMA 传输内核默认开启(#40278)、旧的 MambaRadixCache 类已移除(#40313)。实测数值及异常说明见上表。
**下一步:**报告 readiness-blocked 并释放该配置族:共享的 smoke 生成器会在多节点 AgentX 评测行中输出 eval-all-concs,而矩阵 schema 禁止该字段,主分支修复前无法启动任何更新镜像作业。

@Klaud-Cold

Copy link
Copy Markdown
Collaborator Author

readiness-blocked · Cleanup pending. Stop and confirm owned runs before closing.

中文

readiness-blocked · 清理待完成。先停止并确认自有运行结束,再关闭 PR。

@Klaud-Cold Klaud-Cold closed this Sep 23, 2026
@Klaud-Cold
Klaud-Cold deleted the klaud/auto-e920fad623895413-abb39e882feee984 branch September 23, 2026 01:11
@Klaud-Cold

Copy link
Copy Markdown
Collaborator Author

readiness-blocked · Repairs: 0 · Runs: 35804907428
All owned runs ended. PR closed; branch deleted for retry.

中文

readiness-blocked · 修复次数:0 · 运行:35804907428
所有自有运行均已结束。PR 已关闭;分支已删除,可重新尝试。

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

Status: Done

Development

Successfully merging this pull request may close these issues.

1 participant