I have this tensor A:
<tf.Tensor: shape=(2, 18), dtype=float32, numpy=
array([[0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1., 1., 1., 1., 1.,
1., 1.],
[0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,
0., 0.]], dtype=float32)>
I want to create a mask to add to another tensor with shape (2, 18, 1000)
. That is, I want to expand each number to 1000 of them, so for example, in tensor A, change each 0 to a dimension of 1000 zeros, and in each 1, change each of them to a dimension of 1000 ones. How could I do it?
Edit
Basically, what I want to do is transform tensor A from shape (2, 18)
to shape (2, 18, 1000)
with those 1000 values being 0 or 1
CodePudding user response:
Simply use tf.expand_dims
import tensorflow as tf
a = tf.constant([[0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1., 1., 1., 1., 1.,
1., 1.],
[0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,
0., 0.]])
b = tf.random.normal(shape=[2, 18, 1000])
c = tf.expand_dims(a, axis=2) b
c.shape
# TensorShape([2, 18, 1000])
About broadcasting see, for example, this numpy tutorial