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
- [Fix] disaggregation: Preserve in-flight staging count when reopening a bootstrap room
- [Feat] diffusion: Add LoRA merge mode disabled
- [Perf] [AMD] attention: fused prefix K and FP8 MLA prefill attention via ASM FMHA on gfx1250
- [Test] [NPU] kimi_k3: run the 128k decode prefix case at C4 with repeat_rate=1
- [Perf] [AMD] aiter: Opt unified verify out of the host seq_lens mirror for spec-v2 overlap
- [Feat] hisparse: support DFLASH speculative decoding
- [Fix] speculative: Preserve repetition penalties in DFLASH and DSPARK
- [Bug] Qwen3, Qwen3-VL, Qwen3-VL-MoE, Qwen3-Omni-MoE, Qwen3-ASR and Qwen3 rerankers ignore checkpoint FP8 KV-cache scales
- [Bug] NVFP4 Marlin path (default on SM80-SM90) decodes E4M3 subnormal block scales as 0 with bfloat16 activations
- [Fix] marlin: Decode E4M3 subnormal NVFP4 block scales exactly with bfloat16 activations
- Docs
- Python not yet supported