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
- [Bugfix] Whisper prompt conditioning: use the real `<|startofprev|>` token
- [Bug][XPU]: AutoRound int4 models silently emit garbage under concurrent requests; ARK WOQ backend is auto-selected with no opt-out
- [Performance]: Best practices for preventing GPU OOM in production
- [Bugfix][Distributed] shm_broadcast: copy out-of-band buffers at enqueue to fix zero-copy send race
- [Bugfix][Core] Name the parameter and shapes on weight shape mismatch
- [Performance] MooncakeConnector emits one RDMA descriptor per (layer, block) — but with the unified paged KV layout a block's layers are contiguous; page-level descriptors give up to 19x faster PD transfers
- [Bug]: `bfloat16` Mamba cache dtype crashes `selective_scan_fwd` on NVIDIA L40S
- [Usage]: performance degradation 0.26 vs 0.27 for Qwen3.5 122B A10 on 4xrtx 3090
- [Bug]: extract_hidden_states overestimates DeepSeek-V4 KV cache memory
- [Bugfix][NIXL] Fix PUSH heartbeat failing to renew KV leases
- Docs
- Python not yet supported