I would like to convert an integer tensor, a_dec=tf.constant([2,7], dtype=tf.int32) to a binary rank-2 tensor such as: a_bin=tf.constant([[0,1,0],[1,1,1]], dtype=tf.int32).
Is there any efficient way to do it?
Add a comment
|
2 Answers
If you know the size n of your vector, what about:
a_bin = tf.mod(tf.bitwise.right_shift(tf.expand_dims(a_dec,1), tf.range(n)), 2)
I tried the solution from guik, but failed when converting 1 and 4, like
a = tf.constant([1,7], dtype=tf.int32)
b = tf.mod(tf.bitwise.right_shift(tf.expand_dims(a,1), tf.range(3)), 2)
Actually, the bits are reversed here, so:
b = tf.reverse(b, axis = [-1])
Comments
