Method forward

model/loss.py:119–142

(self, logits, target,
                use_hierar=False,
                is_multi=False,
                *argvs)

                    loss_type, LossType.str()))

    def forward(self, logits, target,
                use_hierar=False,
                is_multi=False,
                *argvs):
        device = logits.device
        if use_hierar:
            assert self.loss_type in [LossType.BCE_WITH_LOGITS,
                                      LossType.SIGMOID_FOCAL_CROSS_ENTROPY]
            if not is_multi:
                target = torch.eye(self.label_size)[target].to(device)
            hierar_penalty, hierar_paras, hierar_relations = argvs[0:3]
            return self.criterion(logits, target) + \
                   hierar_penalty * self.cal_recursive_regularize(hierar_paras,
                                                                  hierar_relations,
                                                                  device)
        else:
            if is_multi:
                assert self.loss_type in [LossType.BCE_WITH_LOGITS,
                                          LossType.SIGMOID_FOCAL_CROSS_ENTROPY]
            else:
                if self.loss_type not in [LossType.SOFTMAX_CROSS_ENTROPY,
                                          LossType.SOFTMAX_FOCAL_CROSS_ENTROPY]:
                    target = torch.eye(self.label_size)[target].to(device)
            return self.criterion(logits, target)

    def cal_recursive_regularize(self, paras, hierar_relations, device="cpu"):
        """ Only support hierarchical text classification with BCELoss