Skip to content

Pull requests: sgl-project/sglang

New pull request New
4,173 Open 25,142 Closed
4,173 Open 25,142 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Reviews
Filter by reviews
No reviews Review required Approved review Changes requested
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

[VLM] Retire abandoned encoder DP requests run-ci
#36996 opened Aug 29, 2026 by mickqian Collaborator Loading…
1
[Speculative] Add bounded compact DFlash KV cache for PD serving jit-kernel memory-pool speculative-decoding
#36995 opened Aug 29, 2026 by Destination2020 Loading…
5 tasks done
2
[Diffusion] Rollout API: filter trajectory latents to the requested step-index diffusion SGLang Diffusion
#36994 opened Aug 29, 2026 by Rockdu Contributor • Draft
5 tasks
[MoE][Ada] Add Qwen4-Exp FP8 Triton configs for NVIDIA L20
#36993 opened Aug 29, 2026 by positive666 Loading…
5 tasks done
[VLM] Drain timed-out EPD encoder health probes run-ci
#36992 opened Aug 29, 2026 by mickqian Collaborator Loading…
1
[Diffusion] Add exact component precision overrides diffusion SGLang Diffusion documentation Improvements or additions to documentation run-ci
#36991 opened Aug 29, 2026 by mickqian Collaborator Loading…
1
Feat[DCP] Support decode context parallelism for DSA models with the flashmla_kv backend deepseek jit-kernel memory-pool
#36990 opened Aug 29, 2026 by Zhiy-Zhang Contributor Loading…
5 tasks
fix(dcp): GLM-5.3-Flash norope c-KV sharding under DCP (Hopper/tilelang only) deepseek hicache Hierarchical Caching for SGLang jit-kernel memory-pool Multi-modal multi-modal language model npu quant LLM Quantization unified-radix-cache
#36989 opened Aug 29, 2026 by junliu-mde Contributor • Draft
1
[VLM] Retire aborted disaggregated prefill results run-ci
#36988 opened Aug 29, 2026 by mickqian Collaborator Loading…
[Docs] Replace stale diffusion compatibility matrix diffusion SGLang Diffusion documentation Improvements or additions to documentation run-ci
#36987 opened Aug 29, 2026 by mickqian Collaborator Loading…
1
feat(rust-server): support OpenAI embedding requests dependencies Pull requests that update a dependency file
#36986 opened Aug 29, 2026 by qiuhuiming Loading…
3 of 5 tasks
test: re-enable FlashInfer per-token NVFP4 coverage blackwell SM100/SM120
#36985 opened Aug 29, 2026 by zianglih Contributor Loading…
5 tasks done
[Fix] Balance NVTX ranges when forward raises
#36984 opened Aug 29, 2026 by pengpengyi92 Loading…
4 tasks done
1
[VLM] Reject malformed images at the decode boundary npu run-ci
#36983 opened Aug 29, 2026 by mickqian Collaborator Loading…
1
[misc] Move cache_protected_len and swa_evict_floor into ReqKvInfo documentation Improvements or additions to documentation unified-radix-cache
#36982 opened Aug 29, 2026 by hnyls2002 Collaborator Loading…
[VLM] Validate EPD embedding layouts before scheduling run-ci
#36980 opened Aug 29, 2026 by mickqian Collaborator Loading…
[Test] Score GPQA through sgl-eval documentation Improvements or additions to documentation
#36979 opened Aug 29, 2026 by hnyls2002 Collaborator Loading…
4
[Spec][Ngram] Expand SAM match length beyond max_trie_depth documentation Improvements or additions to documentation jit-kernel speculative-decoding
#36978 opened Aug 29, 2026 by shliba123 Loading…
perf(attention): reuse TRTLLM MHA split metadata across layers blackwell SM100/SM120
#36976 opened Aug 29, 2026 by YAMY1234 Collaborator • Draft
3 of 5 tasks
config: the lazy imports that buy nothing become eager deepseek diffusion SGLang Diffusion documentation Improvements or additions to documentation hicache Hierarchical Caching for SGLang jit-kernel lora Multi-modal multi-modal language model npu quant LLM Quantization speculative-decoding unified-radix-cache
#36975 opened Aug 29, 2026 by ch-wan Collaborator Loading…
3 of 5 tasks
1
config: the dead record parameters go hicache Hierarchical Caching for SGLang unified-radix-cache
#36974 opened Aug 29, 2026 by ch-wan Collaborator Loading…
3 of 5 tasks
1
config: six more runtime readers ask the bags amd deepseek documentation Improvements or additions to documentation hicache Hierarchical Caching for SGLang model-gateway quant LLM Quantization sgl-kernel speculative-decoding unified-radix-cache
#36973 opened Aug 29, 2026 by ch-wan Collaborator Loading…
3 of 5 tasks
6
config: the resolution callbacks into the record go to zero hicache Hierarchical Caching for SGLang lora Multi-modal multi-modal language model npu speculative-decoding
#36972 opened Aug 29, 2026 by ch-wan Collaborator Loading…
3 of 5 tasks
3
[VLM] Make EPD send completion idempotent run-ci
#36971 opened Aug 29, 2026 by mickqian Collaborator Loading…
perf(gdn): select ReplaySSM verify loop unrolling by shape bypass-fastfail jit-kernel run-ci
#36970 opened Aug 29, 2026 by YAMY1234 Collaborator Loading…
3 of 5 tasks
2
ProTip! Exclude everything labeled bug with -label:bug.