This repository is an example for creating CMake-based pytorch CUDA extension. It's modified from pytorch extension example and scikit-build example.
To build this repository, install essential requirements and then execute python setup.py build
. If you want to use CUDA in custom location (for example your library is installed from conda install cudatoolkit-dev -c conda-forge
), you can give hint to CMake by defining CMake definition CMAKE_CUDA_COMPILER
.