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] Avoid FP8 descales for FA4 BF16 KV decode
- [Bugfix] Bind GPU processes within the CPUs the process was granted
- [Fix] dsv4 encoding: open the assistant turn after inline system messages
- [Fix] Exclude prefill padding from MLA FP8 quantization scales
- [MLX] Support frequency, presence, and repetition penalties
- [Fix] Aggregate all flush-cache worker responses
- [Fix][DSA] Use RAGGED top-k transform for mixed FP8 sparse prefill
- [Bug] Kimi-K3 cookbook image (sglang 0.5.16) hangs in the first TP16 decode step on 4 nodes x 4 GB200
- [Bug] Kimi-K3 TP16 on GB200 (dev 2026-09-11): 'K3 fused KDA decode disabled: unexpected conv/A_log/dt_bias layout' on every layer
- [Bug] qwen38flashnext cookbook image predates #38851: start-up crash in the QSA KV-extraction kernel once max_running_requests reaches ~4000 (e.g. with --disable-radix-cache)
- Docs
- Python not yet supported