No CUDA runtime is found, using CUDA_HOME='/usr/local/cuda'

[mic2112]

Hey @jinfagang,

I'm trying to run and compile DCNv2, on the latest version of pytorch .i.e. 1.12 with a cuda version of 11.1. I get the following warning when running build:

and testgpu.py returns with the following output:

Can you help, please? TIA