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
- [diffusion][xpu] Feed row-strided q/k/v to XPU flash attention without copies
- [PD] Support unequal TP with unified hybrid MLA memory
- [AMD] Recalibrate test_quark_mxfp4 GSM8K thresholds to observed MI355X noise
- [Bug] GLM-5.3-Flash on SM120: fa4 attention backend crashes at CUDA-graph capture (hybrid extend reshape) — triton is the only working backend
- [Bug] DeepSeek-V4-Pro decode ~5% slower at concurrency 1 on GB300 after #39704
- Add yue2
- [Bug] [Responses API] Replayed Codex turn is split into three assistant blocks (phase check in _merge_consecutive_assistant_messages), so the model learns to end its turn before the tool call
- [Rust] Complete gRPC parity with the existing Rust frontend
- [Bug] Glm4MoeDetector changes string-typed argument values that look like JSON ("true" → "True", "1.50" → "1.5")
- [Bug] deepseekv31 detector (streaming): greedy name regex drops a call and attaches the next call's arguments to it
- Docs
- Python not yet supported