Method forward

model/vision_transformer.py:524–544

(self, x)

                nn.init.constant_(m.bias, 0)

    def forward(self, x):
        probability = []
        if self.shared_projected_layer is None:
            attribute_embedding = []
            for classifier in self.classifier_list:
                projected = classifier[0](x)
                _attribute_embedding = classifier[1](projected)
                attribute_embedding.append(_attribute_embedding.detach().clone())
                activate = classifier[2](_attribute_embedding)
                logit = classifier[3](activate)
                probability.append(self._log_softmax(logit))
            attribute_embedding = torch.cat(attribute_embedding, dim=1)
        else:
            attribute_embedding = self.shared_projected_layer(x)
            for classifier in self.classifier_list:
                logit = classifier(attribute_embedding)
                probability.append(self._log_softmax(logit))
        if self.training == True:
            return probability
        else:
            return attribute_embedding


class Attribute_BN_Classifier(nn.Module):