sglang
https://github.com/sgl-project/sglang
Python
SGLang is a high-performance serving framework for large language models and multimodal models.
Triage Issues!
When you volunteer to triage issues, you'll receive an email each day with a link to an open issue that needs help in this project. You'll also receive instructions on how to triage issues.
Triage Docs!
Receive a documented method or class from your favorite GitHub repos in your inbox every day. If you're really pro, receive undocumented methods or classes and supercharge your commit history.
Python not yet supported1 Subscribers
Add a CodeTriage badge to sglang
Help out
- Issues
- [Perf] [NVIDIA] custom-all-reduce: Enable 2shot_push in SM107 TP4/TP8/TP16 heuristic
- [Fix] megamoe: Count the DP-attention chunk once in the token budget check
- [Fix] swa: Start the SWA KV pool at its first layer under pipeline parallelism
- Optimize Kimi K3 NVFP4 prefill and decode
- [Bug] Xid 13 "CTA Not Present" in HiCache kvcacheio transfer kernels on Blackwell (B20Z) — 7 crashes across 4 machines / 7 GPU instances under agentic load
- [Feat] [NPU] multimodal: support mm_feature_transport=npu_ipc
- [Feat] r3: Reuse first-computed routed experts when a session re-prefills
- [Feat] [NVIDIA] megakernels: DSA / MLA binding view, EAGLE draft capture checks, early draft embedding sharing
- [Perf] speculative: opt-in step CUDA graph and fused verify / relay paths for EAGLE v2 chains
- [Feat] [NPU] Quant: support Qwen3.6-35B-A3B-FP8 which is block FP8 MOE.
- Docs
- Python not yet supported