When I have changed the set(USE_CUDA ON) in ‘config.cmake’ file in build directory of TVM then it is giving this error. What to do? I need to use TVM with CUDA as a backend, it working properly with LLVM.
Do you also have USE_CUDNN set to on? try disabling that if you just need to use CUDA without relying on cuDNN libraries; otherwise you need to also install cuDNN.