TensorFlow1.x 中的张量转换为 numpy array 的方法及用途实例

一、将 TensorFlow 中的张量 Tensor 转换为 numpy array 的方法实例

>>> import tensorflow as tf
>>> data_tensor = tf.constant([1,2,3,4,5,6], shape=[2,3], dtype=tf.float32)
>>> data_tensor
<tf.Tensor 'Const:0' shape=(2, 3) dtype=float32>
>>> with tf.Session() as sess:
# 方法一：
...     print("sess.run(tensor): {}".format(sess.run(data_tensor)))
# 方法二：
...     print("tensor.eval(session=sess): {}".format(data_tensor.eval(session=sess)))
... 
# 打印结果：
sess.run(tensor): [[1. 2. 3.]
 [4. 5. 6.]]
tensor.eval(session=sess): [[1. 2. 3.]
 [4. 5. 6.]]

二、数据张量转换为 numpy 数组的应用实例

将包含指定数据的 tensor 输入给 Session 的 run() 方法的 feed_dict 参数，但是需要注意 feed_dict 参数不能接收 tensor，所以需要将 tensor 进行转换为 numpy array 形式

>>> import tensorflow as tf
>>> truncated_tensor = tf.random.truncated_normal([2,3],dtype=tf.float32)
>>> truncated_tensor
<tf.Tensor 'truncated_normal:0' shape=(2, 3) dtype=float32>
>>> X_holder = tf.placeholder(dtype=tf.float32, shape=[2,3])
>>> result = tf.split(X_holder,[1,2],-1)
>>> result
[<tf.Tensor 'split:0' shape=(2, 1) dtype=float32>, <tf.Tensor 'split:1' shape=(2, 2) dtype=float32>]
>>> with tf.Session() as sess:
...     data = sess.run(truncated_tensor)
...     matrix1, matrix2 = sess.run(fetches=result, feed_dict={X_holder:data})
...     print("matrix1: {}".format(matrix1))
...     print("matrix2: {}".format(matrix2))
... 
matrix1: [[ 0.44706684]
 [-1.1646993 ]]
matrix2: [[-1.026939    0.25133875]
 [ 0.3521515  -0.21427773]]

错误方式举例：Session 的 run() 方法的参数 feed_dict 不能接收 tensor 类型

>>> truncated_tensor = tf.random.truncated_normal([2,3],dtype=tf.float32)
>>> truncated_tensor
<tf.Tensor 'truncated_normal:0' shape=(2, 3) dtype=float32>
>>> X_holder = tf.placeholder(dtype=tf.float32, shape=[2,3])
>>> result = tf.split(X_holder,[1,2],-1)
>>> result
[<tf.Tensor 'split:0' shape=(2, 1) dtype=float32>, <tf.Tensor 'split:1' shape=(2, 2) dtype=float32>]
>>> with tf.Session() as sess:
...     matrix1, matrix2 = sess.run(fetches=result, feed_dict={X_holder:truncated_tensor})
...     print("matrix1")
...     print("matrix2")
... 
Traceback (most recent call last):
  File "<stdin>", line 2, in <module>
  File "/******/Anaconda/anaconda3/lib/python3.7/site-packages/tensorflow/python/client/session.py", line 950, in run
    run_metadata_ptr)
  File "/******/Anaconda/anaconda3/lib/python3.7/site-packages/tensorflow/python/client/session.py", line 1124, in _run
    'feed with key ' + str(feed) + '.')
TypeError: The value of a feed cannot be a tf.Tensor object. Acceptable feed values include Python scalars, strings, lists, numpy ndarrays, or TensorHandles. For reference, the tensor object was Tensor("truncated_normal:0", shape=(2, 3), dtype=float32) which was passed to the feed with key Tensor("Placeholder:0", shape=(2, 3), dtype=float32).

（PS：根据报错提示可知， tf.Tensor 不能直接传入给参数 feed_dict，该参数仅支持：Python 标量、字符串、列表，numpy n维数组，或者 TensorHandles 等类型）

三、numpy 数组转换为张量 tensor 的方法实例

>>> import numpy as np
>>> data = np.random.random([2,3])
>>> data
array([[0.02510445, 0.11428815, 0.11885889],
       [0.34326366, 0.69007324, 0.70780292]])
>>> data_tensor = tf.convert_to_tensor(data)
>>> data_tensor
<tf.Tensor 'Const_2:0' shape=(2, 3) dtype=float64>
>>> sess = tf.Session()
>>> data_tensor.eval(session=sess)
array([[0.02510445, 0.11428815, 0.11885889],
       [0.34326366, 0.69007324, 0.70780292]])
>>> sess.close()