CODEWITHSUNDEEP
tensorflow
Bruno KM
Bruno KM

Reputation: 768

TensorFlow equivalent of numpy.all()

As stated in the title, is there a TensorFlow equivalent of the numpy.all() function to check if all the values in a bool tensor are True? What is the best way to implement such a check?

Upvotes: 11

Views: 5000

Answers (3)

David Bacelj
David Bacelj

Reputation: 182

You could use tf.experimental.numpy.all in tf 2.4

x = tf.constant([False, False])
tf.experimental.numpy.all(x)

Upvotes: 1

Miriam Farber
Miriam Farber

Reputation: 19634

Use tf.reduce_all, as follows:

import tensorflow as tf
a=tf.constant([True,False,True,True],dtype=tf.bool)
res=tf.reduce_all(a)
sess=tf.InteractiveSession()
res.eval()

This returns False.

On the other hand, this returns True:

import tensorflow as tf
a=tf.constant([True,True,True,True],dtype=tf.bool)
res=tf.reduce_all(a)
sess=tf.InteractiveSession()
res.eval()

Upvotes: 11

Bruno KM
Bruno KM

Reputation: 768

One way of solving this problem would be to do:

def all(bool_tensor):
    bool_tensor = tf.cast(bool_tensor, tf.float32)
    all_true = tf.equal(tf.reduce_mean(bool_tensor), 1.0)
    return all_true

However, it's not a TensorFlow dedicated funciton. Just a workaround.

Upvotes: 0

Related Questions