ensorflow: tf.stack和tf.unstack的区别，实例解释

将两个N维张量列表沿着axis轴组合成一个n+1维的张量，例如下面tensor(2,3)与tensor1(2,3),一个y(2,2,3)

import tensorflow as tf
tensor=[[1,2,3],[4,5,6]]
tensor2=[[10,20,30],[40,50,60]]
y=tf.stack([tensor,tensor2])
y2=tf.stack([tensor,tensor2],axis=1)
print(y)#Tensor("stack_1:0", shape=(2, 2, 3), dtype=int32)
with tf.Session() as sess:
    print(sess.run(y))
#    [[[ 1  2  3] [ 4  5  6]]
# [[10 20 30] [40 50 60]]]
    print(sess.run(y2))
#[[[ 1  2  3][10 20 30]]
# [[ 4  5  6][40 50 60]]]

将输入的值按照指定行或者列进行拆分，并输出含有num个元素的列表list（axis=0）表示按照行拆分，axis=1按照列拆分，Num是输出list的个数，必须与预计输出的个数相等，否则会报错

import tensorflow as tf
tensor=[[1,2,3],[4,5,6]]
y=tf.unstack([tensor])
y2=tf.unstack([tensor],axis=1)
print(y)#<tf.Tensor 'unstack:0' shape=(2, 3) dtype=int32>]
with tf.Session() as sess:
    print(sess.run(y))
#[array([[1, 2, 3],[4, 5, 6]])
    print(sess.run(y2))
#array([[1, 2, 3]]), array([[4, 5, 6]])

转载自：https://blog.csdn.net/Dian1pei2xiao3/article/details/90214179