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
- [ROCm] mxfp4 MoE: TRITON_UNFUSED is never auto-selected, leaving pre-CDNA3 (gfx90a) with no working backend
- [ROCm][Perf] Fuse act_and_mul into the AITER MXFP4 GEMM's activation quant
- [Feature]: [ROCm][AITER] Track relu2 (non-gated) MoE activation enablement (Tier 3, vLLM-side integration)
- [RFC][Core] Add a model metadata provider hook for plugins
- [Bug]: GLM-5.3-Flash with HiSparse has no compatible KV cache layout (LBHNC vs BLHNC)
- [Bug] VLLM_BATCH_INVARIANT=1 is not batch-invariant for a pooling (cross-encoder) model on sm_120 — GEMMs fall to the cuBLAS-workspace branch
- [Bug]: DFlash DCP slot mapping uses kernel block size for DCP ownership
- [MM][CG] Support ViT full CUDA graph for BLIP-2
- [Transformers backend] Migrate ModernBERT embedding, reranker and token classification models
- [Hisparse] Support HiSparse for QSA
- Docs
- Python not yet supported