Merge and Split
-
tf.concat
-
tf.split
-
tf.stack
-
tf.unstack
1. tf.concat
原有维度concat，如果新维度concat，则stack
-
concat：拼接的维度可以不同，其他维度必须相同。 concat([3, 35, 8], [2, 35, 8]) -> [5, 35, 8]
-
stack：所有维度必须相同。 stack([3, 35, 8], [3, 35, 8]) -> [2, 3, 35, 8]
import tensorflow as tf
# [classes, students, scores]
a = tf.ones([4, 35, 8])
b = tf.ones([2, 35, 8])
# class维度concat
c = tf.concat([a, b], axis=0)
c.shape
TensorShape([6, 35, 8])
# [classes, students, scores]
a = tf.ones([4, 30, 8])
b = tf.ones([4, 5, 8])
# students维度concat
tf.concat([a, b], axis=1).shape
TensorShape([4, 35, 8])
2. tf.stack
stack: create new dim
会创建一个新的维度，设计新的schools概念，在新的维度上表达。
[classes, students, scores] -> [schools, classes, students, scores] 新建schools维度并concat
# [classes, students, scores]
a = tf.ones([4, 35, 8])
b = tf.ones([4, 35, 8])
tf.stack([a, b], axis=0).shape
TensorShape([2, 4, 35, 8])
tf.stack([a, b], axis=3).shape
TensorShape([4, 35, 8, 2])
3. tf.split vs unstack
e = tf.ones([2, 4, 35, 8])
res = tf.unstack(e, axis=3)
len(res), res[0].shape
(8, TensorShape([2, 4, 35]))
res = tf.split(e, axis=3, num_or_size_splits=2)
len(res), res[0].shape
(2, TensorShape([2, 4, 35, 4]))
res = tf.split(e, axis=3, num_or_size_splits=[2, 2, 4])
len(res), res[0].shape, res[2].shape
(3, TensorShape([2, 4, 35, 2]), TensorShape([2, 4, 35, 4]))