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] hisparse: reject unvalidated speculative algorithms by name
- [Perf] [XPU] graph: Support full prefill graph for GDN/Mamba and Gemma models
- [Fix] HiCache: Wait for decode prefetch before publishing prefix length
- [Fix] deepseek_v4: Capture decoder-tail graphs at startup, never on a step a DP rank skips
- [Perf]cosmos3: Explore opt-in BF16 DiT CUDA Graphs
- [Perf] [CPU] graph: Move SWA index translation into Intel AMX attention kernels and prune masked prefix loads
- [Fix] [AMD] all-reduce: Check quick reduce before custom AR so ROCM_QUICK_REDUCE_QUANTIZATION takes effect
- [Perf] [AMD] dsa: Use a spill-free BF16 tile for the triton sparse-MLA split-K decode on gfx950
- [Perf] [AMD] kda: Row-streamed chunked verify for ReplaySSM ring mode on gfx950
- [Perf] [AMD] deepseek-v4.1: Reuse the low-ratio identity extend fill across the layers of a ratio
- Docs
- Python not yet supported