AttributeError: 'Model' object has no attribute 'predict_classes'

Models in keras, both Sequential and Functional, provide following predictions
functions: 

• predict()
• predict_on_batches()
• predict_step()

Apart from this, Sequential model provides two addtional methods: 

• predict_classes()
• predict_proba()

These functions directly predict the resultant class or class probability, 
without needing any conversion like numpy.argmax()
However, predict_classes() and predict_proba() are not supported by 
functional model and you will get following error if you try to implement it on
functional model: 
AttributeError: 'Model' object has no attribute 'predict_classes'