pytorch批量loss

        loss_char =torch.zeros(1).cuda()

        for j in range(len(restored)):
            loss_char=loss_char+criterion_char(restored[j], target)