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]: `POST /reset_prefix_cache?reset_running_requests=true` returns HTTP 500 while an OffloadingConnector load is in flight
- [Performance][NIXL] BLHNC MLA caches lost whole-row registration in #53781 and now need more NIXL descriptors than LBHNC
- [Bug]: AOT compile artifacts fail to load with mode `DYNAMO_TRACE_ONCE`
- [Bugfix][Frontend] Treat empty/whitespace arguments as complete JSON in Anthropic streaming
- [Perf] Keep a 512-token floor on the default CUDA graph capture size
- [Bug][ROCm][MoRIIO] A decode whose notify port is already in use starts and reports healthy without its notify listener, so its WRITE requests hang
- [Bugfix][ROCm] Fix CUDA-only FlashAttention import in MiMo audio encoder
- [Perf][KV Connector] Optimize Mooncake contiguous block grouping
- [Bug]: `kv_cache_dtype="fp8"` auto-selects the FLASHINFER attention backend without a usable FlashInfer JIT and crashes instead of falling back to TRITON_ATTN
- [Spec Decode] Generic reduced-vocabulary get_top_tokens() for DraftModelSpeculator
- Docs
- Python not yet supported