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
- [AMD]perf(mamba): avoid temporary extend state copies in Mamba/GDN backend
- [Bug][NPU][Diffusion] Missing ffprobe in the community image causes completed MiniMax-H3 T2VA outputs to be deleted
- [MiniMax-M3] Enable FlashInfer TRT-LLM MXFP8 MoE and optimize SwiGLU/MSA prefill
- [Bug] DeepSeek-V4 DSpark draft crashes on SM120 because topk=192 misses FlashInfer decode dispatch
- [Spec][DSpark] Clamp per-request verify window to context/generation budget (fix #33454)
- [Bug] It is defined as int32 in SGLang, but the kernel within `ascend_backend` reads it as int64, leading to incorrect memory‑address access.
- [ROCm] Fix HiCache host-pool allocator: hipHostRegister's device pointer is not the host VA
- [PD] Make the mooncake KV transfer wait on the prefill forward that wrote the pages
- [Bug]
- Page-split kernel uses int32 byte-offset math, wraps at ~3.67M tokens/rank on large SM120 KV pools
- Docs
- Python not yet supported