请求大家帮助解答知识蒸馏的程序

distill = Distilling(Student_model, Teacher_model, 2, 0.9)
distill.compile(optimizer='adam',
loss=tf.keras.losses.CategoricalCrossentropy(from_logits=False))

callback = [keras.callbacks.EarlyStopping(patience=20, restore_best_weights=True)]

distill.fit(train_images, train_labels, epochs=500, validation_data=(test_images, test_labels), callbacks=callback)
在distill.compile(optimizer='adam',
loss=tf.keras.losses.CategoricalCrossentropy(from_logits=False))增添 metrics=['accuracy']修改成为
distill.compile(optimizer='adam',
loss=tf.keras.losses.CategoricalCrossentropy(from_logits=False),
metrics=['accuracy'])
但是metrics怎么输出呀？
我试了
def test_step(self, data):
x, y = data
softmax = keras.layers.Softmax()
logits = self.student_model(x)
loss_value = self.compiled_loss(y, softmax(logits))
#auc = self.compiled_metrics.update_state(y, softmax(logits))
#return {'loss':loss_value,'accuracy':auc}
但是报错，恳请大家帮助解答？
十分感谢大家