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
- [HiCache] Size explicit host pools against one snapshot taken by every rank
- [Cookbook] Document BF16 mixed-chunk latency tuning for Qwen3-VL on B300
- [RFC] Use reasoning effort as an output-length hint for DP routing and retraction
- [Bugfix] Fix raw-token limits for bigram prefix matching
- [Bug] Qwen3.8-27B compressed-tensors W4A16 (int4, symmetric, group 128) serves with degraded quality on 0.5.19-0.5.21 (wikitext PPL 9.98 vs 6.05 on vLLM, same checkpoint)
- [HiCache] Budget declared DSA index host pools and keep mismatched MTP draft rows separate
- [Model Loader] Let InstantTensor load DeepSeek-V4.1 with the Engram host table
- [DeepSeek-V4.1] Speed up Engram host-table loading with parallel positioned reads
- [DSv4.1] Speed up prefill with encoder SWA bounded replay by checking the request-window history once per layout, without a host sync
- [DSV4.1] Fix Torch prefill indexer request-pool rows under CP
- Docs
- Python not yet supported