In libTorch, how to set 'None' value in autograd backward?

In python, if we dont want pass gradient in auto backward
we can just return like this: return None, None, a_varianble, None

But How to do this in c++?