Tensorflow--取tensorf指定列的操作方式

时间：2021-05-24 08:51:25|栏目：Python代码|点击：次

我就废话不多说了，大家还是直接看代码吧~

In [1]: import os
In [2]: os.environ["CUDA_VISIBLE_DEVICES"] = "0"
In [3]: import tensorflow as tf
In [4]:sess =tf.Session()
In [5]: input = tf.constant([[[1,2,3],[4,5,6],[7,8,9]],[[10,11,12],[13,14,15],[1
 ...: 6,17,18]]])
 
In [6]: input.get_shape()
Out[6]: TensorShape([Dimension(2), Dimension(3), Dimension(3)])
 
In [7]: input_2 = input[:,:,2]
 
In [8]: print(sess.run(input_2))
[[ 3 6 9]
 [12 15 18]]
 
In [9]: input_2 = input[:,:,0:2]
 
In [10]: print(sess.run(input_2))
[[[ 1 2]
 [ 4 5]
 [ 7 8]]
 
 [[10 11]
 [13 14]
 [16 17]]]
 
In [11]: input = tf.constant([[[[1,2,3],[4,5,6],[7,8,9]],[[10,11,12],[13,14,15],
 ...: [16,17,18]]]])
 
In [12]: input.get_shape()
Out[12]: TensorShape([Dimension(1), Dimension(2), Dimension(3), Dimension(3)])
 
In [13]: input_2 = input[:,:,2]
 
In [14]: print(sess.run(input_2))
[[[ 7 8 9]
 [16 17 18]]]
 
In [15]: input_2 = input[:,:,:,2]
 
In [16]: print(sess.run(input_2))
[[[ 3 6 9]
 [12 15 18]]]

补充知识：TensorFlow 训练过程中获取某个Tensor值；只有conv1和bn1存在NAN

1. 在训练过程中,获取某个参数Tensor的值:

获取所有Tensor的name: