model_CNN_LSTM.py 文件源码

def forward(self, x):
        # print("fffff",x)
        embed = self.embed(x)

        # CNN
        cnn_x = embed
        cnn_x = torch.transpose(cnn_x, 0, 1)
        cnn_x = cnn_x.unsqueeze(1)
        cnn_x = [F.relu(conv(cnn_x)).squeeze(3) for conv in self.convs1]  # [(N,Co,W), ...]*len(Ks)
        cnn_x = [F.max_pool1d(i, i.size(2)).squeeze(2) for i in cnn_x]  # [(N,Co), ...]*len(Ks)
        cnn_x = torch.cat(cnn_x, 1)
        cnn_x = self.dropout(cnn_x)

        # LSTM
        lstm_x = embed.view(len(x), embed.size(1), -1)
        lstm_out, self.hidden = self.lstm(lstm_x, self.hidden)
        lstm_out = torch.transpose(lstm_out, 0, 1)
        lstm_out = torch.transpose(lstm_out, 1, 2)
        # lstm_out = F.tanh(lstm_out)
        lstm_out = F.max_pool1d(lstm_out, lstm_out.size(2)).squeeze(2)

        # CNN and LSTM cat
        cnn_x = torch.transpose(cnn_x, 0, 1)
        lstm_out = torch.transpose(lstm_out, 0, 1)
        cnn_lstm_out = torch.cat((cnn_x, lstm_out), 0)
        cnn_lstm_out = torch.transpose(cnn_lstm_out, 0, 1)

        # linear
        cnn_lstm_out = self.hidden2label1(F.tanh(cnn_lstm_out))
        cnn_lstm_out = self.hidden2label2(F.tanh(cnn_lstm_out))

        # output
        logit = cnn_lstm_out
        return logit