Nsryjdtyk

I am trying to build a multi-input multi-output model using keras functional api and I am following their code but I got that error:

ValueError: A Concatenate layer requires inputs with matching
shapes except for the concat axis. Got inputs shapes: [(None, 50),
(None, 50, 1)]

I have skipped the Embedding layer, here is the code:

def build_model(self):

    main_input = Input(shape=(self.seq_len, 1), name='main_input')

    print(main_input.shape)

    # seq_len = 50

    # A LSTM will transform the vector sequence into a single vector,

    # containing information about the entire sequence

    lstm_out = LSTM(self.seq_len,input_shape=(self.seq_len,1) )(main_input)

    self.auxiliary_output = Dense(1, activation='sigmoid', name='aux_output')(lstm_out)



    auxiliary_input = Input(shape=(self.seq_len,1), name='aux_input')

    print(auxiliary_input.shape)

    x = concatenate([lstm_out, auxiliary_input])



    # We stack a deep densely-connected network on top

    x = Dense(64, activation='relu')(x)

    x = Dense(64, activation='relu')(x)

    x = Dense(64, activation='relu')(x)



    # And finally we add the main logistic regression layer

    main_output = Dense(1, activation='sigmoid', name='main_output')(x)

    self.model = Model(inputs=[main_input, auxiliary_input], outputs=[main_output, auxiliary_output])

    print(self.model.summary())

    self.model.compile(optimizer='rmsprop', loss='binary_crossentropy',

              loss_weights=[1., 0.2])

I got that error in the concatenation step, although printing the shape of both layers are (?,50,1).
I do not know exactly why I got this, and what is the exact error in the input_shape of the first layer and why it does not give me the same shape as it should be using print(main_input.shape), and how to solve it ?

UPDATE:

I found a solution for the error by changing the shape of the second input layer

auxiliary_input = Input(shape=(self.seq_len,), name='aux_input')

so now they can concatenate smoothly, but still not clear to me why ?

For the second input, you specified before the bug that,

input_shape = (50,1)# seq_length=50

This means final shape is:

(None,50,1)

Now, when the first input passes through LSTM , since you didn't specify return_sequences=True it will return a tensor of shape (batch_size, units) viz. (None, 50) which you are concatenating with the above mentioned (None, 50, 1)

Your error went away because you changed the input shape for the second input as (50,) so the final shape becomes (None,50) which matches with output of LSTM and hence it concatenated smoothly