tf.print() vs python print vs tensor.eval()

The native Python print() statement will be called when the graph is build the first time. Check this out:

a = tf.placeholder(shape = None, dtype = tf.int32)
b = tf.placeholder(shape = None, dtype = tf.int32)
print("a is ", a, " while b is ", b)
c = tf.add(a, b)
with tf.Session() as sess:
   print(sess.run(c, feed_dict = {
      a: 1,
      b: 2
   }))
print(sess.run(c, feed_dict = {
   a: 3,
   b: 1
}))

By exectuing this code block, the output is:

# a is Tensor("Placeholder:0", dtype = int32) while b is Tensor("Placeholder_1:0", dtype = int32)
# 3
# 4

On the other hand, let us see tf.print():

a = tf.placeholder(shape = None, dtype = tf.int32)
b = tf.placeholder(shape = None, dtype = tf.int32)
print_op = tf.print("a is ", a, " while b is ", b)
with tf.control_dependencies([print_op]):
   c = tf.add(a, b)

with tf.Session() as sess:
   print(sess.run(c, feed_dict = {
      a: 1,
      b: 2
   }))
print(sess.run(c, feed_dict = {
   a: 3,
   b: 1
}))

No. The Python print is not called when you call sess.run() later.
If you want to print when you call sess.run() then you can use tf.print.
To print out multiple tensor values in a graph, you should use sess.run() after opening tf.Session(). Sample code is below.

t = tf.constant(42.0)
u = tf.constant(37.0)
pt = tf.print(t)
pu = tf.print(u)
with sess.as_default():
   sess.run([pt, pu])

t = tf.constant(42.0)
u = tf.constant(37.0)
pt = tf.print(t)
pu = tf.print(u)
with sess.as_default():
   sess.run([pt, pu])

42
37

Suggestion : 2

Last updated 2022-06-28 UTC.

View aliases

Compat aliases for migration

See Migration guide for more details.

tf.compat.v1.print

tf.print( *
   inputs, ** kwargs
)

tensor = tf.range(10)
tf.print(tensor, output_stream = sys.stderr)

Multi-input usage:

tensor = tf.range(10)
tf.print("tensors:", tensor, {
   2: tensor * 2
}, output_stream = sys.stdout)

Usage in a tf.function:

@tf.function
def f():
   tensor = tf.range(10)
tf.print(tensor, output_stream = sys.stderr)
return tensor

range_tensor = f()

In graphs manually created outside of tf.function, this method returns the created TF operator that prints the data. To make sure the operator runs, users need to pass the produced op to tf.compat.v1.Session's run method, or to use the op as a control dependency for executed ops by specifying with tf.compat.v1.control_dependencies([print_op]).

  tf.compat.v1.disable_v2_behavior() #
  for TF1 compatibility only

  sess = tf.compat.v1.Session()
  with sess.as_default():
     tensor = tf.range(10)
  print_op = tf.print("tensors:", tensor, {
        2: tensor * 2
     },
     output_stream = sys.stdout)
  with tf.control_dependencies([print_op]):
     tripled_tensor = tensor * 3

  sess.run(tripled_tensor)

tensor = tf.range(10)
tf.print(tensor, output_stream = sys.stderr)

Multi-input usage:

tensor = tf.range(10)
tf.print("tensors:", tensor, {
   2: tensor * 2
}, output_stream = sys.stdout)

Changing the input separator:

tensor_a = tf.range(2)
tensor_b = tensor_a * 2
tf.print(tensor_a, tensor_b, output_stream = sys.stderr, sep = ',')

Suggestion : 3

No. The Python print is not called when you call sess.run() later. If you want to print when you call sess.run() then you can use tf.print. ,The native Python print() statement will be called when the graph is build the first time. Check this out:,To print out multiple tensor values in a graph, you should use sess.run() after opening tf.Session(). Sample code is below. ,So, according to the output below, we can see that if we add the dependency that the tf.print op must be run whenever c is run, we get to see the output we want to:

a = tf.placeholder(shape = None, dtype = tf.int32)
b = tf.placeholder(shape = None, dtype = tf.int32)
print("a is ", a, " while b is ", b)
c = tf.add(a, b)
with tf.Session() as sess:
   print(sess.run(c, feed_dict = {
      a: 1,
      b: 2
   }))
print(sess.run(c, feed_dict = {
   a: 3,
   b: 1
}))

tf.print( *
   inputs, ** kwargs
)

Print

import tensorflow as tf

a = tf.placeholder(shape = None, dtype = tf.int32)
b = tf.placeholder(shape = None, dtype = tf.int32)
print("a is ", a, " while b is ", b)
c = tf.add(a, b)
with tf.Session() as sess:
   print(sess.run(c, feed_dict = {
      a: 1,
      b: 2
   }))
print(sess.run(c, feed_dict = {
   a: 3,
   b: 1
}))

Suggestion : 4

Last Updated : 22 Apr, 2022

Syntax:

tf.print(value, verbose)

Output:

Tensor
dtype: string
rank: 2
shape: [2, 2]
values: [
   ['geeks', 'for'],
   ['geeks', 'website']
]

Suggestion : 5

Sometimes we need to fetch and print the value of a TensorFlow variable to guarantee our program is correct., Fetch the value of a TensorFlow variable or a Tensor , Fetch the value of a TensorFlow variable or a Tensor , Using if condition inside the TensorFlow graph with tf.cond

For example, if we have the following program:

import tensorflow as tf
import numpy as np
a = tf.Variable(tf.random_normal([2, 3])) # declare a tensorflow variable
b = tf.random_normal([2, 2]) #declare a tensorflow tensor
init = tf.initialize_all_variables()

if we want to get the value of a or b, the following procedures can be used:

with tf.Session() as sess:
   sess.run(init)
a_value = sess.run(a)
b_value = sess.run(b)
print a_value
print b_value

with tf.Session() as sess:
   sess.run(init)
a_value = a.eval()
b_value = b.eval()
print a_value
print b_value