Hi, Thank u for this great repo. Could you please tell me why using tanh here x = torch.tanh(self.fc_hidden(hidden[0])+self.fc_encoder(encoder_outputs)) Many thanks in advance