Tensorflow基础学习五杂项

split: [[1 2 3]
 [4 5 6]
 [7 8 9]]
split: [[10 11 12]
 [13 14 15]
 [16 17 18]]

x_data_2 = [[1,2,3],[4,5,6],[7,8,9],[10,11,12],[13,14,15],[16,17,18]]
sp = tf.split(x_data_2, [1,2], axis=1)

with tf.Session() as sess:
    for c in sp:
        print("split:",sess.run(c))

split: [[ 1]
 [ 4]
 [ 7]
 [10]
 [13]
 [16]]
split: [[ 2  3]
 [ 5  6]
 [ 8  9]
 [11 12]
 [14 15]
 [17 18]]

常见的用法：

例如：有数据为NHWC--[batch, height, width, channels]

如果想把各个chanels的数据单独拿出来，则可以使用tf.split()