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] qsa: CUDA-graph-safe fused resolve for the QSA K/V host offload (2/N)
- [Perf] qsa: copy only the context rows a prefill reads under the QSA K/V offload
- [Perf] model_executor: Reuse device positions for text-only MRoPE
- [Feat] sampling: --sampling-banned-token-ids masks token ids for the target and EAGLE draft
- [Fix] tokenizer_manager: Leave room for max_new_tokens and reserved tokens when auto-truncating
- [Perf] [NVIDIA] mxfp4: Bound SM90 preprocessing memory
- [Perf] [AMD] dspark: Speed up the DSPARK draft step
- [Perf] radix-cache: Preserve long device prefixes during partial eviction
- [Feat] [AMD] moe: Run AITER MegaMoEGfx1250 with the mori dispatch
- [Feat] pd: Place prefill sessions by projected per-rank token load
- Docs
- Python not yet supported