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 supported51 Subscribers
Add a CodeTriage badge to vllm
Help out
- Issues
- [HelionLinearBackend][4/N] Add HelionINT8ScaledMMLinearKernel
- [Bugfix][Attention] Size FlashInfer paged-KV buffers for local attention
- [Bugfix] Prevent get_open_port() from spinning forever when VLLM_PORT is in the DP reserved range
- [Bug]: /v1/chat/completions/batch accepts stream: true and returns empty output with HTTP 200; tools are silently dropped
- [Bugfix] Reject stream and tools in batch chat completions instead of returning empty output
- [Perf] Use Triton moe backend for tensor fp8 quant scheme on Hopper
- fix(entrypoints): reject stream=True and tools in BatchChatCompletionRequest
- fix(utils): advance next_port when candidate_port is in reserved DP range in get_open_port
- LoRA on a compressed-tensors int4 (W4A16) model: rank-32 all-layer adapters produce weak, non-reproducible outputs; rank-8 partial-coverage adapters work fine (0.17.1–0.24.0, RTX 5090)
- [CUDA] Keep host registration and error drain on one runtime handle
- Docs
- Python not yet supported