I have tried implementing keras.Sequential in two manners having two very distinct performances:
Sequential([...])
model = tf.keras.Sequential([ tf.k