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 supported62 Subscribers
View all SubscribersAdd a CodeTriage badge to vllm
Help out
- Issues
- [Performance]: Waiting requests re-walk their whole prefix-cache hit on every scheduling step
- [Bug]: enable_tower_connector_lora: get_mm_lora_token_counts() under-sizes punica wrappers in 8 models (startup/runtime token_lora_mapping asserts)
- [Bugfix] LLM.__repr__ crashes on non-torch worker models
- [Feat][Quantization] Online linear NVFP4 to MXFP4
- [Bug]: Step-start synchronous .cpu() in attention metadata builders holds the CUDA driver context lock and stalls other threads' CUDA calls
- [Bugfix] Read step-start seq_lens without holding the CUDA driver context lock
- [Bugfix][Rust Frontend] Prevent stalled engines from delaying auto-aborts
- [Perf][Scheduler] Remove redundant pad_spec_decode guard (fix #55798)
- [RFC]: Eager-mode Sequence Parallelism (SP) with Collective+GEMM Overlap (Async-TP)
- [Bugfix][Gemma4] Accept legacy ModelOpt per-expert weight names
- Docs
- Python not yet supported