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] [Apple] load-snapshots: Use an available mmap directory on macOS
- [Fix] disaggregation: propagate Mooncake memory registration failures
- [Bug] Qwen3-VL vision encoder: full-size transient tensors (pos-embed blend `[4, patches, 1152]`, MLP `[patches, 4304]`) OOM memory-tight GPUs on large images; both can be computed in row chunks
- [Fix] multimodal: Decode images on CPU when the PIL image processor is used
- [Fix] utils: Match the CUDA driver's default coredump pipe name
- [Fix] sgl-router: Refuse --route-by-uncached-tokens under reorg routing
- [Fix] mamba: Keep the decode-track ping-pong parity under overlap interleavings
- [Bug] d3 decision checkpoints: the readout load erases the embedding matrix
- [Fix] d3: Keep the decision readout out of the embedding matrix
- [Perf] scheduler: Defer duplicate prompts until KV checkpoint
- Docs
- Python not yet supported