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 supported61 Subscribers
View all SubscribersAdd a CodeTriage badge to vllm
Help out
- Issues
- [Bug]: Ultravox syncs chunk lengths to the host, and the 3000-mel LoRA pad changes odd-length clips
- [Performance]: Qwen2-Audio and AudioFlamingo3 materialize a dense [B, 1, S, S] key-padding mask
- [Bug]: FunASR merges 80 adaptor frames where the prompt length is 10; Cohere, FireRed, and Granite still build full conformer scores
- [Performance]: Batch Whisper's conv stem and CUDA-graph only the fixed full-chunk mel
- [Bugfix][Offload] Free stale pinned copy during prefetch-offload weight loading
- [Bugfix][Model] Route Gemma 4 per-layer embeddings through the CPU weight offloader
- [Performance]: Whisper log-mel runs on the renderer thread instead of the GPU worker
- [Bugfix][Attention] Avoid runtime JIT in Triton unified attention when num_seqs changes
- [Doc] Sync supported_models.md with model registry
- [Model] Fall back to block_configs for heterogeneous Nemotron-H MoE layers
- Docs
- Python not yet supported