How do I train U-Net++ on my custom 3D data?

Hi, I am having a hard time trying to modify the code so that I can train U-Net++ for my 3D data with 1 channel input and 1 channel output. For instance, the shape of input and output would be 16x256x256x1 (Depth,H,W,C).
I do not need any backbone; I want to use the architecture of U-NET++.