Questions about loss function modification #20

Hi, i want to use and modify your code for image inpainting. I have a dataset with thousands of pics. Here are some questions,hope someone can help me: @liuguilin1225
1.if I want to fill a hole in image A with more content from image A and less imaginary content from other images. What should i do? Is it right to modify total loss function? Is so, which part of loss should be modified?
2. If I modify loss function, does it work if I train my own dataset both with the pretrained model this repo provided and the modified loss function?