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
- [Feat] unified-memory: Read each draft step's sliding window on FlashInfer and TRTLLM MHA
- [Fix] attention: Use ModelRunner query width for DSpark FlashAttention drafts
- [Perf] beam_search: Defer token materialization to top-k beams in BeamGroup.finalize()
- [Perf] [AMD] dsa: Run the GLM-5 decode MLA chain as Gluon kernels on gfx950
- [RoadMap] [NPU] NPU Backend Refactor Roadmap
- [Fix] qwen-vl: round video frame indices to match qwen-vl-utils
- Fix issue #46476: DSpark draft query width on FA3/FA4
- Fix issue #46461: HunyuanImage-3 shared MLP execution order
- Fix issue #46447: Kimi-K3 AR fusion buffer capacity with attention-DP
- [Bug] KV-full retraction can drop `new_token_ratio` below its floor and loosen admission
- Docs
- Python not yet supported