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
- [Test] tests: Fix tests broken by recent refactors and pytest helper collection
- [Perf] muse-spark: Enable fused RMSNorm for FP32 residuals
- [Feat] xpu: Consolidate simple XPU enablements
- [Performance] DeepSeek-V4.1-Flash-1017 on 8x H20-3e: SGLang vs DeepInfer #92 measurements and profiling
- [Perf] [AMD] deterministic: Use a fixed-order two-stage all-reduce for large inputs
- [Bug] multimodal_gen: cross-node pure-Ulysses H3 serve (1 GPU/node x 4 nodes) freezes at the first distributed generation - /health never binds with --warmup-mode off, 0% GPU on all ranks, zero errors
- [Bug] FLUX.3 Image: TP/SP ranks pick packed or sequential CFG independently and can hang
- [Perf] paged-experts: exact size pinning
- [Perf] eplb: Fuse expert-count recording into one kernel on CUDA decode graphs
- [Bug][MUSA] fa3: scheduler metadata is computed from the draft backend under speculative decoding, hanging when the radix cache hits
- Docs
- Python not yet supported