如果tensor是用调用tensorflow框架定义的，那么用 tensor_name.shape 即可返回tensorflow 的维数：
>>> import tensorflow as tf
>>> a=tf.constant([
... [[1.0,2.0,3.0,4.0],
... [5.0,6.0,7.0,8.0],
... [8.0,7.0,6.0,5.0],
... [4.0,3.0,2.0,1.0]],
... [[4.0,3.0,2.0,1.0],
... [8.0,7.0,6.0,5.0],
... [1.0,2.0,3.0,4.0],
... [5.0,6.0,7.0,8.0]]
... ])
>>> a.shape
TensorShape([Dimension(2), Dimension(4), Dimension(4)])
也可通过调用 numpy 来返回 tensor 的维数：
>>> import numpy as np
>>> np.shape(a)
TensorShape([Dimension(2), Dimension(4), Dimension(4)])
