GPU Out of Memory deep learning UNET network with tensorflow. How to solve it?

Question

My question is how to calculate the ram memory need to run a deep learning network? I am asking this question because, my training for some network configuration is getting out