No method "predict_classes" in Model

It seems that only Sequential models have the predict_classes method, which means that a model built using the functional API cannot be wrapped using the SKlearn KerasClassifier class. Is there a technical reason this is not implemented?

[ghost]

Here is an example for nlp `

labels = list(set(y))
nb_classes = len(labels) + 1
label2ind = {label: index+1 for index, label in enumerate(labels)}
ind2label = {(index+1): label for index, label in enumerate(labels)}

def encode(y):
    return np.asarray([to_categorical([label2ind[word] for word in sentence], nb_classes) for  sentence in y])

def decode(y):
    y_ind = np.argmax(y, axis=-1)
    return [[ind2label[word] for word in sentence] for sentence in y_ind]`

[spedygiorgio]

It would be nice to add this method to the Model class, in order to give the same funtionalities of the Sequential api