BoyuanJiang/matching-networks-pytorch

Seems fce are not correctly used

Opened this issue · 1 comments

In the forward function of matching_net.py, output of self.g is passed to self.lstm, but the output of self.lstm is not used.

Note the 'output' becomes 'outputs' in 'outputs = self.lstm(output)'

code line 198

        # use fce?
        if self.fce:
            outputs = self.lstm(output)

        # get similarities between support set embeddings and target
        similarites = self.dn(support_set=output[:-1], input_image=output[-1])

Do you suggest then change outputs to output?