How to Iterate Over Tensor In TensorFlow

In this TensorFlow tutorial, I will show you how to iterate over tensors in TensorFlow.

While processing data, I created a custom function to normalize the data and fill in the missing values. I wanted to apply this function to each element of my dataset in tensor format.

For that, I needed to iterate over my dataset, so I used a for-and-while loop and applied the custom function to each element. I successfully preprocessed the dataset.

In this tutorial, you will learn how to create a tensor and iterate over its values individually. Additionally, I have provided solutions to errors that may arise while iterating over the tensor.

What does Iterate Over mean?

Iterate over is a term used in every programming language. It is the way to visit each element of a collection of items sequentially, one by one. The collection can be any iterable data structure, such as a dictionary, set, list, or tensor.

So here, you will learn how to iterate over Tensor. In TensorFlow, a tensor is a multidimensional array that can store different kinds of data.

Lets’ begin

Iterate Over Tensor in TensorFlow using Python Loop

So you can iterate over tensors using the Python For loop. If you want to know how For loop works, visit the Python tutorial For loop vs while loop in Python, and read the For loop section.

To iterate over a tensor, first create a tensor using the tf.constant() function.

import tensorflow as tf

tensor_data = tf.constant(['USA', 'Canada', 'Brazil', 'Austrailia'])

The above code, tensor_data, is created and contains the list of string values.

READ:  Modulenotfounderror no module named ‘tensorflow.keras.utils.np_utils’

Next, use the For loop, as shown in the code below, to iterate over tensor_data and print each string value.

for data in tensor_data:
Iterate Over Tensor in TensorFlow using Python Loop

Look at the output. The For loop printed all the elements of tensor tensor_data. The code ‘for data in tensor_data’ in this line visits each element of tenosr_data one by one, represented by data in the for loop.

Next, let’s iterate over the tensor using a for loop.

Iterate Over Tensor in TensorFlow using While Loop

In the above section, you have used the For loop to iterate over the tensor. Similarly, you can use the While loop of Python.

First, create a tensor using the below code.

tensor_values = tf.constant([23, 56, 98, 55])

Now, initialize the two variables tensor_len and data using the code below.

tensor_len = len(tensor_values)
data = 0

The above two variables, tensor_len, store the length of the tensor_values using the len() function and the data are initialized with a value of 0.

Use the while loop to iterate over the tensor, as shown below.

while data < tensor_len:
Iterate Over Tensor in TensorFlow using While Loop

Look using the While loop; all the elements of the tensor_values are printed on the terminal.

First, the loop checks if the data < tensor_len, the initial value of data is 0, and the tensor_len is 4.

then within the while loop, a statement print(tensor_values[data]), takes the each values from tensor_value variable and prints on the terminal.

Finally, the value variable data is incremented by one in each iteration using data+=1.

While iterating over a tensor, you might encounter errors such as typeerror: cannot iterate over a scalar tensor.

READ:  Python Django round to two decimal places

Here, the error indicates that you can’t iterate over a scalar value. A scalar value is a single value, so you can access one but can’t iterate on it. You can iterate over a collection of values, not on a single value.

Let me show you through code how that error can appear, so use the below code.

import tensorflow as tf


for tensor in tensor:
    print("Iterate tensor:",tensor)
Typeerror cannot iterate over a scalar tensor.

When you execute the above code, the error shows that you have created a tensor using tensor=tf.constant(1), which contains only a single value 1, so you can iterate on it.

The tensor variable must contain a list of values to iterate over the tensor; you can modify the code below.

import tensorflow as tf

for tensor in tensor:
    print("Iterate tensor:",tensor)
Solution to Typeerror cannot iterate over a scalar tensor.

Look now; the error disappears. Here, a tensor is created that contains a list of values [1,7,2,8], and then, using a for loop, each value of a tensor is printed.

You can also get one more error while iterating over a tensor: ‘iterating over a symbolic `tf.tensor` is not allowed.’

To resolve that error, you can use the code below to examine how the Tensorflow function is used.

import tensorflow as tf

tens = tf.random.uniform(shape=(2,50, 60, 400))
result = tf.shape(tens)
tens = [tens[0][m][n] - tf.reduce_mean(tens[0][m][n]) for n in tf.range(result[2]) for m in tf.range(result[1])]

In the above code, we have imported the TensorFlow library and then used the tf.random.uniform() function. Within this function, the random shape is described. After that, the list comprehension method, the for loop and tf.reduce_mean() function is used.

Now, I hope you understand how to iterate over tensors using Python’s for and while loop.


In this TensorFlow tutorial, you learned how to iterate over tensors. You have used the Python For and While loops to iterate over tensors.

READ:  Python Turtle Graphics

Additionally, you learned how to fix the errors while iterating over tensors such as typeerror: cannot iterate over a scalar tensor and ‘iterating over a symbolic `tf.tensor` is not allowed‘.

You may like to read: