For my project, I downloaded cxx11-abi libtorch here, and then pointed my CMakeLists.txt to the installed location via,
find_package(Torch REQUIRED PATHS ${LIBTORCH_PATH})
This call, under its hood, appears to find my system cuda installation, and everything works.
To increase portability, I want to change the project to use the cuda package that is available on conda.
How do I get the above cmake command to find my conda cuda package rather than my system cuda installation?