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 supported49 Subscribers
View all SubscribersAdd a CodeTriage badge to vllm
Help out
- Issues
- [BugFix] Memoize StreamingParserEngine drop-info construction
- [Feature/Design]: UVA weight offload on quantized MoE — pin_memory() power-of-two rounding wastes ~35% host RAM, and MXFP4 process_weights_after_loading silently un-offloads layers via replace_parameter
- [Frontend] Integer token IDs for generate output logprobs (GenerateLogProbs)
- [Bug]: finish_reason is "tool_calls" for a turn truncated at max_tokens, hiding the length stop from clients
- Revert "[Perf] Add fused DFlash2 grouped convolution (#55960)"
- [CI] Raise fault-detection deadline 45s -> 75s in FT E2E test
- [Bug]: CPU W4A16 Whisper fails: cpu_gemm_wna16 rejects 3-D encoder activations; packed k_proj bias is synthesized incorrectly
- [Installation]: v0.30.0 CPU wheels now require manylinux_2_39 (glibc 2.39), breaking installation on Ubuntu 22.04/Rhel9
- [Bugfix] Attach DeepSeek-V4 request tools to existing system/developer messages
- [RFC]: KVPP (KV pipeline parallel, LayerSplit) for vLLM
- Docs
- Python not yet supported