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 supported52 Subscribers
Add a CodeTriage badge to vllm
Help out
- Issues
- [Bugfix][Docker] Add fastokens as optional pip extra and Docker build arg
- [Bug]: `vllm bench sweep serve_workload` crashes with "non-positive request rate" when max throughput < 1 req/s (integer floor/ceil of rate anchors)
- [ROCm] Make aiter fused-qk-rmsnorm capability probe crash-safe
- [RFC]: FlashInfer NVFP4 KV serving on pre-SM100 GPUs
- [Bugfix][Pooling] Preserve cache_salt for chunked embeddings
- [Bugfix][Config] Expand debug dump path before absolutizing
- [Warmup] Tune the LM head GEMM during FlashInfer autotune
- RuntimeError: shape mismatch during KV cache init with EP + DP on MoE model
- fix: bad invalid expert IDs and scaling factor in FlashInfer all-to-all communication integration
- [Bug]: Mistral models with tool_choice=required + streaming generates invalid tool_call_id format
- Docs
- Python not yet supported