Repository navigation
[Klaud Cold] Update Qwen3.8-Flash-Next FP8 H200 SGLang AgentX image to v0.5.21-cu130 / 将 Qwen3.8-Flash-Next FP8 H200 SGLang AgentX 镜像更新至 v0.5.21-cu130 - #3961
Conversation
Pin lmsysorg/sglang:v0.5.21-cu130 by digest and rename the removed --cuda-graph-max-bs alias to --cuda-graph-max-bs-decode. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
Thanks for the contribution!
中文感谢你的贡献！
|
|
Initial attempt · Passed · Run 38401077777 / attempt 1 · 2026-10-13 14:18 UTC
Note: All rows: Δ N/A: dataset/scenario/statistic mismatch.
Note: All rows: Δ N/A: no matched eval baseline. Next: Append the changelog entry and run the final full sweep. 中文初次尝试 · 已通过 · Run 38401077777 / attempt 1 · 2026-10-13 14:18 UTC |
…pdate Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Klaud-Cold
added
the
full-sweep-fail-fast
|
Final full sweep · Cancelled · Run 38412326100 / attempt 1 · 2026-10-13 14:20 UTC
Note: All rows: cancelled. Next: Follow the superseding sweep run 38412393944 on the same head. 中文最终完整 sweep · 已取消 · Run 38412326100 / attempt 1 · 2026-10-13 14:20 UTC |
|
Final full sweep · Passed · Run 38412393944 / attempt 1 · 2026-10-13 16:37 UTC
Note: All rows: Δ N/A: dataset/scenario/statistic mismatch.
Note: All rows: Δ N/A: no matched eval baseline. Next: Mark ready for maintainer review. 中文最终完整 sweep · 已通过 · Run 38412393944 / attempt 1 · 2026-10-13 16:37 UTC |
|
View unofficial run (performance): https://inferencex.semianalysis.com/inference?unofficialRun=38412393944 View unofficial run (accuracy): https://inferencex.semianalysis.com/evaluation?unofficialRun=38412393944 |
|
validated · Repairs: 0 · Runs: 38401077777, 38412326100, 38412393944 中文validated · 修复次数：0 · 运行：38401077777, 38412326100, 38412393944 |
|
/use 38412393944 |
|
Claude finished @Klaud-Cold's task in 14s —— View job LGTM - no blocking issues found |
|
@Klaud-Cold staged run 38412393944. This run remains available across future |
![claude[bot]](https://avatars.githubusercontent.com/in/1236702?s=60&v=4){kind=small}
There was a problem hiding this comment.
Looks good, straightforward image bump + mechanical flag rename.
What was reviewed: the image pin update in both the recipe YAML and nvidia-master.yaml (strings match exactly, sha256 digest is well-formed), the cuda-graph-max-bs → cuda-graph-max-bs-decode rename across all five concurrency overrides (c1/c4/c8/c12/c16) with no leftover old-name references, and the perf-changelog.yaml entry (appended at tail, single PR link, matches the config-key).
Extended reasoning...
Three-file config-only diff: a pinned SGLang image bump kept identical between the single-node recipe and nvidia-master.yaml, a uniform launch-arg rename (cuda-graph-max-bs -> cuda-graph-max-bs-decode) applied consistently to all five concurrency override blocks with nothing left stale, and one well-formed perf-changelog entry appended at the file tail citing PR #3961. No code logic, auth, or security-sensitive paths are touched, the change follows the repo's documented image-bump/changelog conventions, and the bug hunt reported no findings, so this is simple and mechanical enough to approve without further human review.
This branch has not been deployed
1 participant
Goal: Update SGLang image from
lmsysorg/sglang:qwen38flashnexttolmsysorg/sglang:v0.5.21-cu130@sha256:b1259f3ea3275f66237c498ea388919729018bc9f01c3d638391e06e2cf3f469.Baseline: 2026-08-27 ·
lmsysorg/sglang:qwen38flashnextAgentX · TP4/EP4 · Mean latency · Sources: API 1, API 2, API 3
中文
**目标：**将 SGLang 镜像从
lmsysorg/sglang:qwen38flashnext更新为lmsysorg/sglang:v0.5.21-cu130@sha256:b1259f3ea3275f66237c498ea388919729018bc9f01c3d638391e06e2cf3f469。**基线：**2026-08-27 ·
lmsysorg/sglang:qwen38flashnextAgentX · TP4/EP4 · 平均延迟 · 来源： API 1, API 2, API 3；数值及异常说明见上表。
AI model disclosure
Prepared autonomously by Klaud Cold using
claude-opus-5-5(edits, upstream source comparison, dispatch and reporting). No delegated agents.🤖 Generated with Claude Code