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
View all SubscribersAdd a CodeTriage badge to vllm
Help out
- Issues
- [Bug]: XPU TP>1 consumes host RAM equal to total VRAM; workers are never isolated to their own device
- [ROCm] Tests confirmed skipped due to num_devices/num_gpus mismatch (nightly-log-verified)
- [Bug]: Out-of-bounds attrIdxs in the C++ batch memcpy path (cache_kernels.cu)
- DFlash2 Ampere: callable hf_overrides blocks online FP8; Marlin packs DFlash K=0; need Dynamo-safe W8 draft path
- [Bug]: Qwen3.5 GDN — ~1.9x long-prefill throughput regression and ~1.15 GiB extra device memory vs 0.1.dev19754
- [RFC]: Using Helion for Selected vLLM CustomOps
- [Bugfix] [Perf] Honor explicit FlashInfer all-reduce size override for ds4-pro-dspark(h200)
- [Bugfix] Fix Exaone input embedding delegation
- [Bug]: v0.28.0 hangs when starting distributed inference with DeepSeek-V4-Pro on 2 nodes × 16 H100 GPUs
- [MoE] Support out-of-tree fp8 MoE backends
- Docs
- Python not yet supported