Keras Functional API: ValueError: could not broadcast input array from shape (75,11) into shape (75,)

Question

I have a model using Keras Functional API with 2 outputs: (CNN and regression) model with 2 outputs:

trying to use predict but getting ValueError: could not broadcast input array from shape (75,11) into shape (75,)

here's the code of training process:

fold_no = 1
    kfold = KFold(n_splits=num_folds, shuffle=True)
    for tr, valid in kfold.split(train_images, train_labels):
        print('------------------------------------------------------------------------')
        print(f'Training for fold {fold_no} ...')
        # Train the model
        history= model.fit(data_images[tr], data_vector[tr],
                                      batch_size=batch_size,
                                      epochs=80,
                                      verbose=verbosity,
                                      validation_split=0.2)

        scores = model.evaluate(data_images[valid], data_vector[valid], verbose=0)

        print(data_labels[valid])
        print(np.argmax(model.predict(data_images[valid]), axis=-1))

        fold_no += 1

printed the shape of:

• data_images[valid]: (75, 5, 28, 3)
• data_vector[valid]: (75, 11)

Please any suggestions to solve this problem will be helpful, I need a way to take results from the trained model. thanks in advance

Answer 1

I found the reason for the error: predict for 2 outputs model returns an array with the results of prediction for both model outputs. to reach one of them can use an index, in my example: model.predict(data_images)[0] for classification and model.predict(data_images)[1] for regression