How to compile ADOP for an specific CUDA ARCH

[MarcoG5]

First I would like to thank you for your greate work and making it public!

I have already build and run it sucessfully in my own machine and it works fine. And I want to build a version to support both NVIDIA Tesla T4 (sm_75) and Tesla V100 (sm_70) to train adop. I wonder how to compile a version to support both sm_75 and sm_70. Or how to compile a version to support sm_70 on a sm_75 device?