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 supported59 Subscribers
View all SubscribersAdd a CodeTriage badge to vllm
Help out
- Issues
- [Bugfix][Env] Lowercase VLLM_FLOAT32_MATMUL_PRECISION before torch sees it
- [ROCm][Perf] Use the Triton fp32 router GEMM up to 128 tokens for small routers on gfx950
- [Build] Prune docs and CI config from the sdist
- [Benchmark] Add --random-mm-video-encode-fps to random-mm
- [Perf][Kernel] Fuse RMSNorm + block FP8 quant for fp32 (Gemma-style) norm weights
- [Benchmark] Add synthetic audio support to random-mm
- [Feature]: vllm.Render for completions, Anthropic Messages, and Responses, with raw-JSON input and optional client-schema output
- [Build][CUDA] Raise CUDA compiler version gates for cooperative_topk (12.8) and DeepGEMM (12.9)
- [Performance]: Gemma 4 full-attention layers run the Triton kernel for the whole prefill whenever the batch carries image/video spans
- [Performance]: Large CPU->GPU KV offload loads stall the worker thread inside cuMemcpyBatchAsync
- Docs
- Python not yet supported