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
- [KV Offload][OBJ] Pass the accelerated option to the NIXL OBJ backend to enable S3 over RDMA
- [KVConnector] Reuse the Mooncake PD TransferEngine in the store connector
- [Bugfix][Realtime] Ignore a final commit when no transcription is running
- [ROCm][Kernel] Pick the split of `top_k_per_row_decode` per row on the device
- [Kernel][Attention] Segment sliding windows in split-K; split-K for non-causal speculative rows (supersedes #61760)
- [Perf][MiniMax-M3] Quantize the routed MoE input while the router GEMM runs
- [Feature]: CUDA 12 wheels for v0.33.x (torch 2.14.0 has cu126 only), and three stale CUDA gates that break CUDA 12.6 source builds
- [Perf] Make detokenizer output-text appends O(1) per token
- [torch.compile] Only flag `_buffers.update` in the cudagraph buffer assignment check
- [KV Connector][Spec Decode] Support lookahead block hashes in remaining KV connectors
- Docs
- Python not yet supported