How to assign a value to a TensorFlow variable in Python

Today in this tutorial, we will learn about how to assign a value to a variable in TensorFlow with Python code example. Variables in a model are those that hold and update the parameters during the execution of the program/code.

First of all, you shall know how to declare a variable in TensorFlow Python. While creating a variable you need to pass a Tensor as variable store these tensors as their initial value.
Here is an example:

biases = tf.Variable(tf.zeros([150]), name="biases")

Then the step is to initialize the variable. To do so we will use the tf.initialize_all_variables() function. This step is essential as it is important that all the variables are initialized and ready to use before other operations are executed.
Let’s see through an example:

biases = tf.Variable(tf.zeros([150]), name="biases")

# Initialize the variables.
init_op = tf.initialize_all_variables()

# Later, when launching the model
with tf.Session() as session:

# Run the init operation.
session.run(init_op)

# Can use the model now

The next step will be saving the variable you have created in order to make sure that you can use them in the future. An easy way to do so to use the tf.train.Saver() to create a saver.

v1 = tf.Variable(..., name="v1")

# Initialize the variables.
init_op = tf.initialize_all_variables()

# Save the variables.
saver = tf.train.Saver()

Assign a value to a variable:

Now that you know how to create, initialize, and save a variable in TensorFlow in Python, I will tell you how to assign the values.

We use the variable_name.assign() function TensorFlow to assign a value to variable.

Now using different versions of TensorFlow can have some issues as the syntaxes are different in the TensorFlow 1 to TensorFlow 2. So I shall teach you about both of the syntaxes.

Let’s begin with the older version of TensorFlow first.

import tensorflow as tf
import numpy as np

x = tf.Variable(0)
init = tf.initialize_all_variables()
sessiom = tf.InteractiveSession()
session.run(init)

print(x.eval())

assign_op = x.assign(1)
sess.run(assign_op)  
print(x.eval())

Output:

0
1

The x.assign(1) does not exactly assign 1 to x, rather it creates a function tf.Operation() is created by it. This should actually be done by the programmer explicitly to update the values of variables.

So how do we do it in TensorFlow 2.0? Well, it is not that much of a difference for this, let’s see how:

import tensorflow as tf
import numpy as np

x = tf.Variable(0)
init = tf.initialize_all_variables()
session = tf.InteractiveSession()
session.run(init)

print(x.numpy())

x.assign(1)
print(x.numpy())

Output:

0
1

So this how you can assign value to a variable in TensorFlow in Python.
Hope you enjoyed learning with me in this session. Have a good day and happy learning.

Leave a Reply

Your email address will not be published. Required fields are marked *