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
- [Bug]: Any KV connector makes pipeline-parallel decode 50-90% slower on Model Runner V2 (all ranks reply, reply-ring writer spins holding the GIL)
- [Bug]: MooncakeStoreConnector crashes EngineCore when a request is preempted by `reset_prefix_cache(reset_running_requests=True)` and re-admitted in the next step
- [Bug][ROCm]: ROCm nightly images not published since 2026-09-25
- [ROCm][Perf][Kimi-K3] Accumulate the latent-MoE up-projection in aiter a16w16 atomic
- [Bug]: QSA indexer uses cooperative_topk on sm_110 (Jetson AGX Thor) and fails with "cluster misconfiguration"
- [ROCM][KVConnector][MoRIIO] Add WRITE-mode consumer timeout
- [Bug]: V2 speculative prefill can change Qwen3-4B's first greedy token through RMSNorm autotune configuration
- [Bugfix][collect-env] Report the cuDNN runtime version
- [feat] Introduce platform-owned RoPE implementations
- [RFC]: Meta-device operator capture: see which operators a model runs on this platform, without weights or device memory
- Docs
- Python not yet supported