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]: FA4 ignores `num_splits` on SM90 in v0.30.0 causing upto 48% slower decode
- [Test][Metrics] Cover MRV2 CUDA graph runtime statistics
- [Spec Decode][Qwen4Exp] Run GDN RecoverSSM verify in the fused CUDA MTP kernel
- [Bugfix][Attention] Reserve the sparse indexer's DCP top-k merge scratch through the workspace manager
- [Bugfix][MRV2] Measure the CUDA graph pool for the memory estimate, not the capture's free-memory delta
- [KV Connector] Send a valid single-token request to the prefiller in the LMCache proxy
- [Frontend][RL] Verify weights against a baseline in finish_weight_update
- [Refactor] KV offload copy backend interface
- [Bug]: SIGTERM during engine startup is swallowed by zmq_socket_ctx; API server hangs until VLLM_ENGINE_READY_TIMEOUT_S
- [Refactor] Use early-return pattern in BlockHashToBlockMap.get_one_block
- Docs
- Python not yet supported