Please explain why you are setting the training parameter only for BatchNormalization
. The test code explicitly manages the learning state via tf.keras.backend.set_learning_phase(True)
and tf.keras.backend.set_learning_phase(False)
Please explain why you are setting the training parameter only for BatchNormalization
. The test code explicitly manages the learning state via tf.keras.backend.set_learning_phase(True)
and tf.keras.backend.set_learning_phase(False)