ValueError: too many values to unpack (expected 2)

[HanGuangXin]

File "train_kd.py", line 331, in train t_emb, t_logits = teacher(images) ValueError: too many values to unpack (expected 2)

[WangYZ1608]

Officially implemented resnet or other models, you need to make a little modification to extract the embedding features. As follows：

def forward_impl(self, x: Tensor, embed):
        ...
        ...
        x = self.avgpool(x)
        emb_fea = torch.flatten(x, 1)
        logits = self.fc(emb_fea)
        if embed:
            return emb_fea, logits
        else:
            return logits

[HanGuangXin]

Yes, I guess so.