Bidirectional RNN Argument Issue

[MdAsifKhan]

Hi, I am getting the following error while using bidirectional RNN with 1DCNN.

error: File "build/bdist.linux-x86_64/egg/keras/layers/containers.py", line 68, in add File "/home/user1/keras-master/seya/seya/layers/recurrent.py", line 55, in set_previous self.forward.set_previous(layer, connection_map) TypeError: set_previous() takes exactly 2 arguments (3 given)

My model is:

forward_lstm = LSTM(input_dim=32, output_dim=32, return_sequences=True) backward_lstm = LSTM(input_dim=32, output_dim=32, return_sequences=True) brnn = Bidirectional(forward=forward_lstm, backward=backward_lstm, return_sequences=True)
model = Sequential() model.add(Convolution1D(input_dim=10, input_length=100, nb_filter=32, filter_length=7, border_mode="valid", activation="relu", subsample_length=1)) model.add(MaxPooling1D(pool_length=3, stride=3)) model.add(Dropout(0.2)) model.add(brnn)) model.add(Dropout(0.5)) model.add(Flatten()) model.add(Dense(output_dim=128)) model.add(Activation('relu')) model.add(Dense(output_dim=1)) model.add(Activation('sigmoid'))

[EderSantana]

Hi, thanks for the issue @MdAsifKhan! Keras had a few API changes and set_previous was probably modified. I'll look at that soon. But if you are able fix the problem, please make us a PR and I'd love to review your contribution. Best, -eder

[zzukun]

The function 'set_previous' was modified to {def set_previous(self, layer, reset_weights=True)} in Keras, I just delete the third parameter and it works.

[EderSantana]

@zzukun cool! would you make a PR?