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
- [Bugfix][Benchmark] Preserve custom audio without a tokenizer
- [Perf][GDN] Static mixed-batch metadata and grids for the GDN prefill path (CUDA graph prep 1/2)
- [Perf][Compile] Enable norm+quant fusion by default for NVFP4 checkpoints
- [Perf][Attention] FA2: faster short-query attention over long cached contexts
- [Observability] Add request tracing for pooling entrypoints.
- [RFC]: Layerwise KV Pooling and P/D Disaggregation for High-Hit-Rate Agent Workloads
- [Rust Frontend] Pass pulled object storage model dir to the Rust frontend
- [UMBP][9/10] Add lazy GPU cache offloading
- [Test][UMBP][10/10] Add lazy offload and token restore coverage
- [Bug]: fused_recurrent KDA path loads the per-key gate g without a mask, resulting in out-of-bounds reads
- Docs
- Python not yet supported