NVIDIA /cudnn-frontend
cuDNN 前端是 NVIDIA 现代开源的 cuDNN 库入口,以及一个不断增长的高性能开源内核集合。
Python996

grouped-gemm 话题在 GitHub 上没有填写简介,这里按 Star 数汇总最相关的热门仓库。
cuDNN 前端是 NVIDIA 现代开源的 cuDNN 库入口,以及一个不断增长的高性能开源内核集合。
Workload-aware persistent expert-tile scheduling for irregular MoE inference kernels on NVIDIA Blackwell GPUs
High-performance grouped matrix multiplication for fine-grained and ultra-fine-grained MoE workloads.
4-bit GPU kernels that compute directly on packed Mixture-of-Experts weights, for training and inference.
Hand-built MoE expert parallelism: routing-skew capture, fused grouped-GEMM Triton kernels, 2-all2all dispatch/combine, overlap ablations — proving by counter-example why DeepEP exists.