cuDNN Frontend is NVIDIA's modern, open-source entry point to the cuDNN library and a growing collection of high-performance open-source kernels.
-
Updated
Jun 30, 2026 - Python
cuDNN Frontend is NVIDIA's modern, open-source entry point to the cuDNN library and a growing collection of high-performance open-source kernels.
Native C++/CUDA and CuTe DSL kernel library for edge MoE inference: flash decode, sync-free GroupGEMM+SwiGLU, head_dim=512 attention
Add a description, image, and links to the grouped-gemm topic page so that developers can more easily learn about it.
To associate your repository with the grouped-gemm topic, visit your repo's landing page and select "manage topics."