Source compile error at the very begining

When I just try to compile Pytorch from source, I encounter this issue(Python 3.6.3, GCC 7.2.0, CUDA Version 8.0.61, CUDNN v5):


does anyone know how to fix it?

If possible, could you downgrade you gcc to something like 4.8
I am not sure that this version of gcc is supported by nvcc in CUDA 8.0 ?