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
- [Fix] function_call: infer argument types from full JSON schema in step3/minicpm5
- [Fix] cuda-graph: Refuse the metadata glue graph when the attention backend plans on the host
- [Perf] pp: Send decode activations before CPU result processing
- [Fix] deepseek_v4: floor amax before the division in the V4.1 fp8 KV scale
- [Bug] DeepSeek-V4.1-Flash: every inference request crashes with IndexError at logits_processor.py:692 on sglang 26396bf566 (empty hidden_states from V4.1 forward)
- [Fix] responses: Align protocol for logprobs, background/stream, and input validation
- [Feat] pd: Add input-only prefix retention on decode
- [Fix] qwen3_coder: Preserve streamed text and tool argument order
- [Feat] Humming: Unify MoE input processing
- [Fix] dsa: Cap the MQA-logits budget by host memory on integrated GPUs
- Docs
- Python not yet supported