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
- [Perf] [AMD] dsv41: write prefill wo_a token-major and hand wo_b its MXFP8 on gfx950
- [Perf] [AMD] Qwen3.8: Serve QSA decode with AITER's fused FlyDSL sparse decode on gfx950
- [Tracking] Frozen-KV MTP (Gemma 4) performance on non-Blackwell GPUs (Triton path)
- [Fix] model_loader: Fix KV scale name remapping and pad_loaded_weight device/integer splitting
- [Fix] sidecar: launch managed sidecars on KV-publishing followers
- [Fix] humming: Free replaced MoE weights after each sublayer transform
- [Perf] [AMD] Kimi-K3: prepare the attention interface for mono fusion
- [Fix] multimodal: Compact CPU tensor views before inline transport
- [Fix] model-loader: Make --weight-loader-drop-cache-after-load effective for mmap loads
- Arbitrary local file read via image_url/video_url/audio_url bypasses the media SSRF allowlist
- Docs
- Python not yet supported