[sglang-bot]

Pins sgl-amd-mono (the monokernels_hip extra) at ca929b7d, up from b10f3b71 (full diff).

• 6777a899 [Fix] glm53_flash: Read the KDA layer cache from the linear-attention backend ([Fix] glm53_flash: Read the KDA layer cache from the linear-attention backend sgl-amd-mono#94)
• 3ad67175 [Fix] deepseek_v41: Pass input_ids to SGLang's Engram so image rows stay in its gate ([Fix] deepseek_v41: Pass input_ids to SGLang's Engram so image rows stay in its gate sgl-amd-mono#65)
• 80396359 [Feat] [MI355X] muse_spark: Whole-layer monokernels for Muse Spark ([Feat] [MI355X] muse_spark: Whole-layer monokernels for Muse Spark sgl-amd-mono#26)
• 796a231a [Perf] [MI355X] deepseek_v41: Push gathered rows in whole 128 B lines ([Perf] [MI355X] deepseek_v41: Push gathered rows in whole 128 B lines sgl-amd-mono#91)
• 517040fb [Chore] maintainer: updates ([Chore] maintainer: updates sgl-amd-mono#96)
• 7bdb7fcc [Perf] [MI355X] glm53: Shared layer and MoE hand-offs, NextN entry, K-split front, indexer launch, MoE expert dedup and weight loads, kernarg reload, LM-head loads ([Perf] [MI355X] glm53: Shared layer and MoE hand-offs, NextN entry, K-split front, indexer launch, MoE expert dedup and weight loads, kernarg reload, LM-head loads sgl-amd-mono#85)
• ea88326a [Perf] [MI355X] deepseek_v41: DSpark step glue in fewer launches, Markov draft sampling, faster K1 and K2 ([Perf] [MI355X] deepseek_v41: DSpark step glue in fewer launches, Markov draft sampling, faster K1 and K2 sgl-amd-mono#95)
• ca929b7d [Perf] [MI355X] glm53: Long-context indexer: two-row scoring from 4 rows, keys straight into registers, the scoring layout ahead, first digits from the scoring ([Perf] [MI355X] glm53: Long-context indexer: two-row scoring from 4 rows, keys straight into registers, the scoring layout ahead, first digits from the scoring sgl-amd-mono#86)

🤖 Generated with GitHub Actions

---

CI States

Latest PR Test (Base): ✅ Run #41798371622
Latest PR Test (Extra): ❌ Run #41798371433
Latest PR Test (AMD ROCm 10): ⏳ Run #41798371610

Tip

Base CI only covers the core cases, and its workload may not exercise your change. Use /rerun-test to run the tests your change touches, or copy this prompt for an agent:

Find the registered tests this PR affects and run them with /rerun-test, following https://github.com/sgl-project/sglang/blob/main/.agents/skills/rerun-test/SKILL.md

[claude]

Claude Code Review

This repository is configured for manual code reviews. Comment @claude review for a one-time review, or @claude review always to subscribe this PR to a review on every future push.

_{Tip: disable this comment in your organization's Code Review settings.}

[BBuf]

[BBuf Review Agent] LGTM. Please wait for AMD validation of the pinned plugin.

[BBuf]

/tag-and-rerun-ci

[BBuf]

[BBuf Review Agent] Could you rebase this onto the latest main? It conflicts with main now.

[BBuf]

[BBuf Review Agent] Closed as superseded: since #45202, main pins sgl-amd-mono at e52803f9, which already contains ca929b7d (sgl-amd-mono#86, the GLM-5.3 work) plus 5 later commits. Please ignore the rebase request above, nothing more is needed here.