[Python] In my Tensorflow model, why is the output of this concatenation the wrong shape? It...

After passing an input sequence to a 1D CNN, I'm trying to combine the output with some engineered features about the sequence. The input sequence is shape (29,1) and the input to the model is (Batch_size, 29, 1). The engineered input is shape (Batch_size, 2). After the CNN the sequence becomes (Batch_size, 29, 512) which I then flatten to (Batch_size, 14848). When I concatenate the two the shape becomes (Batch_size, 14850). Why do I get an error for the Dense layer x7 saying the input its receiving is shape (Batch_size, 1053)?

cnn_output = Activation('relu')(x6)

# --- Combining Branches ---
cnn_flat = Flatten()(cnn_output) # flatten the CNN branch
engineered_input = Input(shape=(num_engineered_features,), name='engineered_input')
combined = Concatenate(axis=1)([cnn_flat, engineered_input])

# --- Final Dense Layers ---
x7 = Dense(64, activation='relu')(combined)


I've confirmed these sizes by printing the sizes of the tensors between the layers.

cnn_output = Activation('relu')(x6)

# --- Combining Branches ---
print(f"Shape of CNN Output before flattening: {cnn_output.shape}.")
cnn_flat = Flatten()(cnn_output) # flatten the CNN branch
print(f"Shape of CNN Output after flattening: {cnn_flat.shape}.")
engineered_input = Input(shape=(num_engineered_features,), name='engineered_input')
print(f"Shape of engineered_input: {engineered_input.shape}.")
combined = Concatenate(axis=1)([cnn_flat, engineered_input])
print(f"Shape of the concatenated tensor: {combined.shape}.")

# --- Final Dense Layers ---
x7 = Dense(64, activation='relu')(combined)


Which outputs:

Shape of CNN Output before flattening: (None, 29, 512).
Shape of CNN Output after flattening: (None, 14848).
Shape of engineered input: (None, 2).
Shape of the concatenated tensor: (None, 14850).


Before giving the following error: Input 0 of layer "dense_1" is incompatible with the layer: expected axis -1 of input shape to have value 14850, but received input with shape (None, 1053)

I've also tried serving an individual batch to the model to ensure it's not a batch sizing issue and I still received this shape mismatch error.

Continue reading...