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] observability: Snapshot STANDALONE metrics before settlement
- [Perf] [AMD] attention: Use the AITER FAv3 varlen kernel on gfx942 as well
- [Fix] disaggregation: Agree on each decode KV transfer's outcome across PP stages
- [Perf] [AMD] kda: Narrow the Kimi-K3 TP8 KDA recurrent tile for small batches on gfx950
- [Perf] [AMD] dsv4: Use the aiter MegaMoE small- and large-input paths
- [Fix] kernels: Hash tensors over 8 GiB in gpu_tensor_hash
- [Feature] diffusion: Add HTTP API key authentication aligned with SRT
- [Fix] dumper: Advance step once per EAGLE speculation round
- [Bug] dumper.step() is advanced per forward, so one EAGLE speculation round spans several dump steps
- [Perf] [NVIDIA] moe: Add H20 block-FP8 Triton configs for Qwen3-30B-A3B
- Docs
- Python not yet supported