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] moe/topk: Use the Triton softmax top-k router for Qwen3-Next (hidden 2048)
- [Fix] dsa: Refuse the trtllm backend at launch on CUDA outside SM10.x
- [TEMP/DO NOT MERGE] NPU: single-case verification for qwen3_6_27b gpqa
- [Fix] diffusion: Check cancellation at stage and sampling boundaries
- [Fix] encoder: Prevent scheduler crash on Encoder failure
- A retracted + re-prefilled request restarts its sampling-penalty and `min_new_tokens` history from zero, silently changing its output
- [Perf] deepseek-v4.1: Use reduce-scatter for CP MoE output
- [Refactor] scheduler: Drop the dead Req.decoded_text and BatchTokenIDOutput.decoded_texts
- [Fix] disagg: Defer aborted prefill release until source reads drain
- [Fix] quantization: Preserve CT MoE zero-point channel order
- Docs
- Python not yet supported