Regarding the backward propagation in the custom loss function

Hi, I am rather new and trying to learn new things from other people's code. I am not able to find examples of code snippets for training with custom loss function in pytorch. Could you suggest me some code snippets/projects from where I may get the idea.