张量tensor的数据与numpy 数据之间的转化与打印

在tensorflow 中一般数据都是用tensor来表示，而在python 中一般是用numpy包，然而有时候需要打印变量的数据，所以下面可以代码：

import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data


def weight_variable(shape):
    initial=tf.truncated_normal(shape,stddev=0.1)
    return initial


if __name__=='__main__':
 '''
    mnist = input_data.read_data_sets("MNIST_data/", one_hot=True)
    x = tf.placeholder(tf.float32, [None, 784],'x')
    y_ = tf.placeholder(tf.float32, [None, 10],'y')
    x_image = tf.reshape(x, [-1, 28, 28, 1])
 '''
    W_conv1 = weight_variable([5, 5, 1, 32])

现在要打印W_conv1变量的值，首先看一下加入print（W_conv1）打印的效果

import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data


def weight_variable(shape):
    initial=tf.truncated_normal(shape,stddev=0.1)
    return initial


if __name__=='__main__':

    '''mnist = input_data.read_data_sets("MNIST_data/", one_hot=True)
    x = tf.placeholder(tf.float32, [None, 784],'x')
    y_ = tf.placeholder(tf.float32, [None, 10],'y')
    x_image = tf.reshape(x, [-1, 28, 28, 1])'''

    W_conv1 = weight_variable([5, 5, 1, 32])
    print (W_conv1)
   # b_conv1 = bias_variable([32])

结果为：Tensor("truncated_normal:0", shape=(5, 5, 1, 32), dtype=float32)，这是张量tensor,要打印这样的变量，需要在session 中，所以加入 with tf.Session() as sess 。程序如下：

import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data


def weight_variable(shape):
    initial=tf.truncated_normal(shape,stddev=0.1)
    return initial


if __name__=='__main__':

    '''mnist = input_data.read_data_sets("MNIST_data/", one_hot=True)
    x = tf.placeholder(tf.float32, [None, 784],'x')
    y_ = tf.placeholder(tf.float32, [None, 10],'y')
    x_image = tf.reshape(x, [-1, 28, 28, 1])'''

    W_conv1 = weight_variable([5, 5, 1, 32])
    with tf.Session() as sess:
            print (sess.run(W_conv1))
   # b_conv1 = bias_variable([32])

打印结果为5*5*1*32,所有复制了部分数据如下：

 [[-0.01138691 -0.06353019  0.12420362  0.14429896  0.01513781
    -0.1557924  -0.11008668 -0.05706662  0.04268187  0.02811845
     0.03623001 -0.15556422 -0.09083539  0.02959536 -0.00855448
     0.09642988 -0.09893788 -0.04005608 -0.02273898 -0.03150641
    -0.00355575  0.120373    0.03304343 -0.04112866  0.02824191
     0.04924054  0.07960307 -0.0673811   0.0637029  -0.10225315
     0.15535429 -0.01115945]]


  [[ 0.1377502  -0.13991357  0.05424013  0.15169595 -0.09851914
     0.0651127   0.05134506  0.0107875   0.01194856  0.00058991
     0.02181729 -0.09146625  0.04761747 -0.06609621  0.03090276
    -0.10576289 -0.00022165  0.06573081 -0.08087479 -0.06913969
     0.05346059 -0.06754622 -0.08353492  0.025491    0.06887309
     0.03608112  0.11575726  0.04321242  0.0215429  -0.07847684
    -0.06896582  0.00389415]]]

通过.eval函数可以把tensor转化为numpy类数据，程序如下：

import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data


def weight_variable(shape):
    initial=tf.truncated_normal(shape,stddev=0.1)
    return initial


if __name__=='__main__':

    '''mnist = input_data.read_data_sets("MNIST_data/", one_hot=True)
    x = tf.placeholder(tf.float32, [None, 784],'x')
    y_ = tf.placeholder(tf.float32, [None, 10],'y')
    x_image = tf.reshape(x, [-1, 28, 28, 1])'''
    sess=tf.Session()
    W_conv1 = weight_variable([5, 5, 1, 32])
    a=W_conv1.eval(session=sess)
    print (a)
   # b_conv1 = bias_variable([32])

通过tf.convert_to_tensor函数可以把numpy转化为tensor 类数据：

import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data


def weight_variable(shape):
    initial=tf.truncated_normal(shape,stddev=0.1)
    return initial


if __name__=='__main__':

    '''mnist = input_data.read_data_sets("MNIST_data/", one_hot=True)
    x = tf.placeholder(tf.float32, [None, 784],'x')
    y_ = tf.placeholder(tf.float32, [None, 10],'y')
    x_image = tf.reshape(x, [-1, 28, 28, 1])'''
    sess=tf.Session()
    W_conv1 = weight_variable([5, 5, 1, 32])
    a=W_conv1.eval(session=sess)
    b=tf.convert_to_tensor(a)
    print (b)
   # b_conv1 = bias_variable([32])