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
- [Refactor] quantization: ask the platform for its quantization config on every device
- [Perf] DeepSeek-V4.1: compile the vision encoder's norm, rotary and gated activation
- [AMD] DeepSeek-V4: compile the FP4 indexer prefill schedule kernels at startup
- [Fix] moe: Use single-writer prefixes in ep_scatter
- [Fix] kimi-k3: Avoid symm-mem deadlock in shared experts
- [Perf] [AMD] spec: Split greedy-verify argmax and prefer aiter tuned bf16 rows on gfx950
- [Perf] [AMD] mla: FP8 prefix K/V pack and absorbed short cached extends for Kimi-K3 prefill on gfx950
- [Perf] [AMD] kimi-k3: Stream the attention-residual bank read for prefill on gfx950
- [Perf] [AMD] kimi-k3: MXFP4 prefill activations, SiTU front epilogue and row-sharded tails on gfx950
- [Perf] [AMD] kda: Kimi-K3 verify tuning, four-step ReplaySSM commit and chunk-major prefill scan on gfx950
- Docs
- Python not yet supported