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 supported43 Subscribers
View all SubscribersAdd a CodeTriage badge to vllm
Help out
- Issues
- [Bug]: Decode Context Parallelism (`--decode-context-parallel-size`) output drift and gibberish in v0.21.0 and latest nightly
- 📝 Integration Proposal: CAJAL — Scientific Paper Model Serving
- Layerwise reload crashes with CUDA illegal memory access on compressed-tensors channel-wise FP8 MoE
- [Bug]: Pipeline Parallelism scheduler does not split sequences into pipeline micro-batches
- Keep first/last n token in high precision for nvfp4 kv cache
- [RFC]: Long-context-optimized Pipeline Parallelism, CPP + Async P2P + Dynamic Chunking
- [Bug]: DeepEP MoE all-to-all backend integration is unusable on Blackwell (SM103 / GB300)
- [Bug]: SimpleCPUOffloadScheduler misses final full block when request finishes in the same scheduler step
- [Feature]: [IR] mm_encoder_attn migration on hold pending FlashInfer workspace support
- [Bug]: TurboQuant _continuation_prefill workspace allocation fails at long context — v0.20.0 regression
- Docs
- Python not yet supported