vllm
https://github.com/vllm-project/vllm
Python
A high-throughput and memory-efficient inference and serving engine for LLMs
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 supported62 Subscribers
Add a CodeTriage badge to vllm
Help out
- Issues
- [Perf][Attention] Split only the sliding window into Triton split-K decode segments
- [Bug]: DP > 1 + MoE expert LoRA: each DP replica applies its own LoRA state to tokens gathered from all replicas (dropped or cross-applied expert delta)
- [Bug]: MiDashengLM-7B generates only out-of-vocab token 151872 (empty text), even for text-only prompts
- [Feature] Support DCP for Gemma 4 sliding-window/full-attention hybrids
- [ROCm][Perf] Sparse MLA: build pure-decode indptr on the host and skip the NoPE q copy
- [Model] Plumb swiglu_limit through Cohere2Moe
- [Model][DSV4.1] Support MXFP4-packed Engram embedding tables
- [Bugfix][Spec Decode] Keep DFlash full-attention layers bidirectional for speculators checkpoints
- [Docs] Document NIXL KV connector metrics aggregation semantics
- [Build] Move extras to pyproject.toml and gate precompiled wheels on [build-system]
- Docs
- Python not yet supported