den = torch.sum(predict.pow(self.p) + target.pow(self.p)) + self.smooth why predict and target need add pow?
