tf.shape()和x.get_shape().as_list()用法

tf.shape()用法：获取张量的大小

打印张量，结果是该张量的类型和形状。

import tensorflow as tf

import numpy as np

a_array = np.array([[1, 2, 3], [4, 5, 6]])
b_list = [[1, 2, 3], [3, 4, 5]]
c_tensor = tf.constant([[1, 2, 3], [4, 5, 6]])
print(a_array)
print(b_list)
print(c_tensor)
# [[1 2 3]
#  [4 5 6]]
# [[1, 2, 3], [3, 4, 5]]
# Tensor("Const:0", shape=(2, 3), dtype=int32)

print(tf.shape(a_array))
print(tf.shape(b_list))
print(tf.shape(c_tensor))
# Tensor("Shape:0", shape=(2,), dtype=int32)
# Tensor("Shape_1:0", shape=(2,), dtype=int32)
# Tensor("Shape_2:0", shape=(2,), dtype=int32)

with tf.Session() as sess:
    print(sess.run(c_tensor))
    # [[1 2 3]
    #  [4 5 6]]

    print(sess.run(tf.shape(a_array)))
    print(sess.run(tf.shape(b_list)))
    print(sess.run(tf.shape(c_tensor)))
    # [2 3]
    # [2 3]
    # [2 3]

x.get_shape().as_list()用法：

x.get_shape()，只有tensor才可以使用这种方法，返回的是一个元组。

import tensorflow as tf

import numpy as np

a_array = np.array([[1, 2, 3], [4, 5, 6]])
b_list = [[1, 2, 3], [3, 4, 5]]
c_tensor = tf.constant([[1, 2, 3], [4, 5, 6]])

print(tf.shape(a_array))
print(tf.shape(a_array).get_shape())
print(tf.shape(a_array).get_shape().as_list())
# Tensor("Shape:0", shape=(2,), dtype=int32)
# (2,)
# [2]

print(tf.shape(b_list))
print(tf.shape(b_list).get_shape())
print(tf.shape(b_list).get_shape().as_list())
# Tensor("Shape_3:0", shape=(2,), dtype=int32)
# (2,)
# [2]

print(c_tensor)
print(c_tensor.get_shape())
print(c_tensor.get_shape().as_list())
# Tensor("Const:0", shape=(2, 3), dtype=int32)
# (2, 3)
# [2, 3]

with tf.Session() as sess:
    print(sess.run(tf.shape(a_array)))
    print(sess.run(tf.shape(b_list)))
    print(sess.run(tf.shape(c_tensor)))
    # [2 3]
    # [2 3]
    # [2 3]

结论：可见只能用于tensor来返回shape，但是是一个元组，需要通过as_list()的操作转换成list。

如果不是元组进行操作的话，会报错！

如果你在上面的代码上添加上a_array.get_shape()会报如下的错误：

    print(a_array.get_shape())
AttributeError: 'numpy.ndarray' object has no attribute 'get_shape'

注意：

第一点：tensor.get_shape()返回的是元组，不能放到sess.run()里面，这个sess.run()里面只能放operation和tensor；否则报错：

# TypeError: Fetch argument TensorShape([Dimension(2)]) has invalid type <class 
# 'tensorflow.python.framework.tensor_shape.TensorShapeV1'>, must be a string or Tensor. (Can not convert a 
# TensorShapeV1 into a Tensor or Operation.) 

第二点：tf.shape(）返回的是一个tensor。要想知道是多少，必须通过sess.run()打印出来。

import tensorflow as tf

import numpy as np

a_array = np.array([[1, 2, 3], [4, 5, 6]])
b_list = [[1, 2, 3], [3, 4, 5]]
c_tensor = tf.constant([[1, 2, 3], [4, 5, 6]])

with tf.Session() as sess:
    a_array_shape = tf.shape(a_array)  # tensor
    print("a_array_shape:", a_array_shape)
    print("a_array_shape:", sess.run(a_array_shape))
    c_tensor_shape = c_tensor.get_shape().as_list()  # list
    print("c_tensor_shape:", c_tensor_shape)
    # a_array_shape: Tensor("Shape:0", shape=(2,), dtype=int32)
    # a_array_shape: [2 3]
    # c_tensor_shape: [2, 3]

    print(sess.run(c_tensor_shape))  # 报错
    # TypeError: Fetch argument 2 has invalid type <class 'int'>, must be a string or Tensor.(Can not convert a int
    # into a Tensor or Operation.)

sess.run()中不是tensor或者是operation，会报错。