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
- [Bug]: Compressed-tensors NVFP4/MXFP4 row-parallel layers load with other groups' block scales when the TP split cuts a scale group (Gemma-4-26B-A4B NVFP4 at TP=8: GSM8K 24% -> 1.4%)
- [ROCm][Model] Add Flint support on AMD GPUs
- [Core] Simplify `prefill_schedule_interval` outside data parallelism
- [Perf][Kernel] Fused MQA+top-k DSA prefill on SM90 with FlashTopK
- [RFC]: P/D disaggregation for continuous RL rollouts (S1): order-free, seconds-long weight-update pauses
- [CPU][recipes] Add Qwen3-VL-4B-Instruct support for Intel Xeon 6
- [Bug]: `compilation_config.backend` accepts arbitrary strings and dies with a bare `AttributeError: module '__main__' has no attribute 'backend'` inside `resolve_obj_by_qualname` — no validation, misleading error
- [Bug][Model] dots.ocr/dots.mocr: vision tower host syncs make the model untestable under VLLM_GPU_SYNC_CHECK
- [Bug]: Single-token benchmark probes inherit generation-length overrides from the main workload
- [Bug]: Invalid `SamplingParams.structured_outputs` type raises `AttributeError` during request validation
- Docs
- Python not yet supported