A question about "l1 loss" between pred outcome and all one tensor in trainer

[YoungQuasimodo]

For code in shading_controlnet_trainer.py,line 247reg_loss = torch.nn.functional.l1_loss(pred_mult_layer, torch.ones_like(pred_mult_layer)) reg_loss += torch.nn.functional.l1_loss(pred_div_layer, torch.ones_like(pred_div_layer))，
why calculating L1 loss between pred outcome and all one tensor, is it to minimize over-exposure artifacts?
Ant it would be greatly appreciated if you could make the dataset and inference code available