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
- [Bug] hicache: Host cache initialization fails in TDX/CC when cudaHostRegister is unavailable
- [Fix] [AMD] quark: Keep MXFP8 shared experts out of the MXFP4 fused slot on dense-first models
- [Test] hicache: Run the unified transfer-lifecycle cases on a fake device for events and streams
- [Perf] [AMD] spec: Sort-free top-p and many-row softmax for HIP verify
- [Perf] flashinfer: Add opt-in FP8 KV dequant for eager prefill
- [Perf] glm: Reduce CP MoE input communication with FP8 transport and fused packing
- [Feat] [Ascend] attention: Use sgl-kernel-npu quantized sparse attention for DSpArk
- [Fix] [NVIDIA] flashinfer_mla: Fall back to FA2 ragged prefill for head dims CUTLASS lacks
- [Perf] [AMD] disaggregation: Cache DCP pack metadata on the gather stream
- [Fix] metrics: Avoid blocking the scheduler on KV metrics backpressure
- Docs
- Python not yet supported