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
- [RL] Return replay outputs through a Mooncake output store
- [Bugfix] Preserve Anthropic streamed tool arguments across formatting whitespace
- [Fix] Reject non-positive server stream intervals
- [MUSA] Add direct AOT RoPE and fused mul-add tests
- test(diffusion): batch performance guard retry subprocesses
- [Fix] Select the earliest completing literal stop in speculative decoding
- [Bug] DeepSeek V4 Pro loading on Lustre takes 95 minutes despite prefetch; limiting tensor-copy workers reduces it to 3.3 minutes
- [Bug] DeepSeek-V4.1 fails at attention backend init on SM12x (GB10): the candidate indexer requires DeepGEMM paged sparse MQA logits, which are SM100-only
- [DSV4.1] Fuse Hopper prefill indexer scoring into one Triton kernel
- [Bug] SM120: DeepSeek-V4.1-Flash decode CUDA graph capture fails with FlashInfer 0.7.0 because `_DECODE_DSV4_DISPATCH` is no longer iterable
- Docs
- Python not yet supported