How to train a neural network twice without re-initialize the model?

Question

let\'s say I have this model:

def mask_layer(tensor):
return layers.Multiply()([tensor, tf.ones([1, 128])])


def get_model():

inp_1 = keras.Input(shape=(64,