TensorFlow's Print is not printing

Question

I am trying to understand some code from a reinforcement learning algorithm. In order to do that I am trying to print the value of a tensor.

I made a simple piece of code to show what I mean.

import tensorflow as tf
from keras import backend as K

x = K.abs(-2.0)
tf.Print(x,[x], 'x')

The goal is to have the value '2' printed(the absolute value of -2). But I only get back the following:

Using TensorFlow backend.

Process finished with exit code 0

Nothing, how can I print the value '2' just like a print('...') statement would do?

Are you trying to print values of Tensors? Could you please specify the type of variable you are trying to print? (Use `print(type(x))` ) — Saket Kumar Singh, Oct 10 '18 at 07:54
Could you edit the question to be 'How to print Tensor value' because that would help others as well. I have answered the question. Hope it helps! — Saket Kumar Singh, Oct 10 '18 at 08:11
You're welcome! Could you mark my answer as correct for the question? — Saket Kumar Singh, Oct 10 '18 at 08:34

score 5 · Accepted Answer · answered Oct 10 '18 at 08:09

If you are using Jupyter Notebook, then tf.Print() so far isn't compatible and would be printing the output to Notebook's server output as is described in the docs

In the tensorflow documentation, here is how Tensors are described:

When writing a TensorFlow program, the main object you manipulate and pass around is the tf.Tensor. A tf.Tensor object represents a partially defined computation that will eventually produce a value.

Hence, you would have to initialize them with a tf.Session() to get their value. To print the value, you eval()

Here is the code you want:

import tensorflow as tf
from keras import backend as K

x= K.abs(-2.0)
with tf.Session() as sess:
    init = tf.global_variables_initializer()
    sess.run(init)
    print(x.eval())

The initializer is important to actually initialize x.

score 1 · Answer 2 · answered Mar 18 '20 at 12:07

To print a tensor in TF 2.0 and above

my_sample = tf.constant([[3,5,2,6], [2,8,3,1], [7,2,8,3]])

With session.run()
with tf.compat.v1.Session() as ses: print(ses.run(my_sample))
One line with eval()
print(tf.keras.backend.eval(my_sample))

score 0 · Answer 3 · answered Apr 09 '20 at 15:04

For learning purpose, sometimes it is convenient to turn on the eager execution. With eager execution enabled, TensorFlow will execute operations immediately. You can then simply use the print or tensorflow.print() to print out the value of your object.

import tensorflow as tf
from keras import backend as K

tf.compat.v1.enable_eager_execution() # enable eager execution

x = K.abs(-2.0)
tf.Print(x,[x], 'x')

see here for more detail. https://www.tensorflow.org/api_docs/python/tf/compat/v1/enable_eager_execution

TensorFlow's Print is not printing

3 Answers3