Below is my network compile statement, which is running fine in TF2.0
model.compile( loss="sparse_categorical_crossentropy", optimizer=a
