Keras: can I use model.predict but not model.predict_generator to predict if I train the model with model.fit_generator

Question

I am new to keras . Now I am going to predict test image groups with the model I trained using model.fit_generator . Can I use model.predict ? Not sure how to use model.predict_generator . And the literature showed the results of these two are different, which one is better?

Thanks a lot!

datagen = ImageDataGenerator(
    enter code here`rotation_range=
    ...
    zoom_range = 0.05) 


model = Sequential()
model.add 
...


model.fit_generator(datagen.flow(train_x, train_y, batch_size=batch_size), 
    steps_per_epoch=train_x.shape[0] // batch_size, 
    epochs=epochs, validation_data=(test_x, test_y))


from sklearn.metrics import roc_auc_score
test_y_hat = model.predict(test_x)
roc_auc_score(test_y, test_y_hat)

Answer 1

Yes you can do that Taking a look at the code for fit_generator() you can see that model does not remember how it was trained. In fact you can use all these APIs ( fit , fit_generator , predict , predict_generator ) on a keras model in any order.