BatchNorm是什么&如何查看keras模型每一层的结果？

似乎数据类型先转为float32为好，除了complex类型的数据。不然可能报错。

>>> xx.shape
(8, 10, 10)

>>> xx2=tf.constant(xx,tf.float32)
>>> inputs=keras.Input(shape=xx.shape[1:],tensor=xx2)
>>> with tf.Session() as sess:
	print(sess.run(inputs))

上面这个是查看最基本的inputs，然而我直接打印BN后的结果出现错误，what's up ?

>>> xx3=keras.layers.BatchNormalization(input_shape=xx.shape[1:])(inputs)
>>> with tf.Session() as sess:
	print(sess.run(xx3))


Traceback (most recent call last):
  File "D:\python\lib\site-packages\tensorflow_core\python\client\session.py", line 1365, in _do_call
    return fn(*args)
  File "D:\python\lib\site-packages\tensorflow_core\python\client\session.py", line 1350, in _run_fn
    target_list, run_metadata)
  File "D:\python\lib\site-packages\tensorflow_core\python\client\session.py", line 1443, in _call_tf_sessionrun
    run_metadata)
tensorflo