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 supported29 Subscribers
Add a CodeTriage badge to vllm
Help out
- Issues
- [NVFP4][Emulation] Fuse NVFP4 weight dequantization with compute in triton kernel for w13/w2 MOE MLP linears
- fix(scheduler): eager KV cache prefetch for waiting queue requests
- Fix memory pointer overflow in Mamba state buffers
- Fix RMS quant fusion for mixed dtype fused add RMSNorm
- [Core] DBO ++: Overlap TP all-reduce with compute
- Fix #44676: Treat <tool_call> as implicit reasoning end in ThinkingBudgetStateHolder
- [Bug]: External DP Load Balancing plus DeepGEMM MegaMoE
- [Performance]: Triton unified attention: sliding-window / chunked KV-tile loop iterates an extra tile when the window start is not tile-aligned
- [DSv4] Pack KV caches into contiguous per-block allocations for DeepSeek V4
- [RFC] Add KVarN: Variance-Normalized KV-Cache Quantization (4-bit K, 2-bit V)
- Docs
- Python not yet supported