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
- [Perf][Attention] DiffKV prefill tiling on SM80: BLOCK_M=128 + KV tile 64 gives 3.0x on 200K-context chunks
- WIP: [Bugfix][Model] Preserve all Kimi-Audio embeddings in batched requests
- WIP: [Bugfix][ROCm] Preserve 64-bit QuickReduce phase offsets
- [Bug]: Severe output degeneration on SM120 with Apertus-70B FP8 in vLLM 0.25.1; v0.22.1 works after rollback
- [Bugfix][Spec Decode] Support dynamic speculative decoding in ngram_gpu (#60981)
- [Docs][Distributed] Document GLOO_SOCKET_IFNAME for docker pipeline parallel deployments (#59109)
- [Test] Make the determinism tests runnable for MLA models
- [Test] Compare NVFP4 batch-invariance logprobs bitwise
- [EPLB] Make migration peer limit configurable
- [Bug]: V2 rejection_sample drops target tokens with ids >= draft vocab width (probabilistic drafts from a narrower draft head)
- Docs
- Python not yet supported