Introduction to the Python Deep Learning Library TensorFlow

Last Updated on December 20, 2019

TensorFlow is a Python library for fast numerical computing created and released by Google.

It is a foundation library that can be used to create Deep Learning models directly or by using wrapper libraries that simplify the process built on top of TensorFlow.

In this post you will discover the TensorFlow library for Deep Learning.

Kick-start your project with my new book Deep Learning With Python, including step-by-step tutorials and the Python source code files for all examples.

Let’s get started.

Introduction to the Python Deep Learning Library TensorFlow

Introduction to the Python Deep Learning Library TensorFlow
Photo by Nicolas Raymond, some rights reserved.

What is TensorFlow?

TensorFlow is an open source library for fast numerical computing.

It was created and is maintained by Google and released under the Apache 2.0 open source license. The API is nominally for the Python programming language, although there is access to the underlying C++ API.

Unlike other numerical libraries intended for use in Deep Learning like Theano, TensorFlow was designed for use both in research and development and in production systems, not least RankBrain in Google search and the fun DeepDream project.

It can run on single CPU systems, GPUs as well as mobile devices and large scale distributed systems of hundreds of machines.

How to Install TensorFlow

Installation of TensorFlow is straightforward if you already have a Python SciPy environment.

TensorFlow works with Python 2.7 and Python 3.3+. You can follow the Download and Setup instructions on the TensorFlow website. Installation is probably simplest via PyPI and specific instructions of the pip command to use for your Linux or Mac OS X platform are on the Download and Setup webpage.

There are also virtualenv and docker images that you can use if you prefer.

To make use of the GPU, only Linux is supported and it requires the Cuda Toolkit.

Your First Examples in TensorFlow

Computation is described in terms of data flow and operations in the structure of a directed graph.

  • Nodes: Nodes perform computation and have zero or more inputs and outputs. Data that moves between nodes are known as tensors, which are multi-dimensional arrays of real values.
  • Edges: The graph defines the flow of data, branching, looping and updates to state. Special edges can be used to synchronize behavior within the graph, for example waiting for computation on a number of inputs to complete.
  • Operation: An operation is a named abstract computation which can take input attributes and produce output attributes. For example, you could define an add or multiply operation.

Computation with TensorFlow

This first example is a modified version of the example on the TensorFlow website. It shows how you can create a session, define constants and perform computation with those constants using the session.

Running this example displays:

Linear Regression with TensorFlow

This next example comes from the introduction on the TensorFlow tutorial.

This examples shows how you can define variables (e.g. W and b) as well as variables that are the result of computation (y).

We get some sense of TensorFlow separates the definition and declaration of the computation from the execution in the session and the calls to run.

Running this example prints the following output:

You can learn more about the mechanics of TensorFlow in the Basic Usage guide.

More Deep Learning Models

Your TensorFlow installation comes with a number of Deep Learning models that you can use and experiment with directly.

Firstly, you need to find out where TensorFlow was installed on your system. For example, you can use the following Python script:

For example, this could be:

Change to this directory and take note of the models subdirectory. Included are a number of deep learning models with tutorial-like comments, such as:

  • Multi-threaded word2vec mini-batched skip-gram model.
  • Multi-threaded word2vec unbatched skip-gram model.
  • CNN for the CIFAR-10 network.
  • Simple, end-to-end, LeNet-5-like convolutional MNIST model example.
  • Sequence-to-sequence model with an attention mechanism.

Also check the examples directory as it contains an example using the MNIST dataset.

There is also an excellent list of tutorials on the main TensorFlow website. They show how to use different network types, different datasets and how to use the framework in various different ways.

Finally, there is the TensorFlow playground where you can experiment with small networks right in your web browser.

Need help with Deep Learning in Python?

Take my free 2-week email course and discover MLPs, CNNs and LSTMs (with code).

Click to sign-up now and also get a free PDF Ebook version of the course.

TensorFlow Resources

More Resources


In this post you discovered the TensorFlow Python library for deep learning.

You learned that it is a library for fast numerical computation, specifically designed for the types of operations that are required in the development and evaluation of large deep learning models.

Do you have any questions about TensorFlow or about this post? Ask your questions in the comments and I will do my best to answer them.

Develop Deep Learning Projects with Python!

Deep Learning with Python

 What If You Could Develop A Network in Minutes

...with just a few lines of Python

Discover how in my new Ebook:
Deep Learning With Python

It covers end-to-end projects on topics like:
Multilayer PerceptronsConvolutional Nets and Recurrent Neural Nets, and more...

Finally Bring Deep Learning To
Your Own Projects

Skip the Academics. Just Results.

See What's Inside

28 Responses to Introduction to the Python Deep Learning Library TensorFlow

  1. Jatin November 29, 2016 at 7:33 pm #

    This was great help.

    Can you post some more tutorials using tensor-flow.


  2. Amit Kumar February 7, 2017 at 9:58 pm #

    print this line is syntactically incorrect. print is a method so should have an opening and closing bracket.

    • Jason Brownlee February 8, 2017 at 9:35 am #

      Thanks Amit, fixed.

    • Asesh April 20, 2017 at 5:49 pm #

      What if it’s Python 2.7? Isn’t the print statement without bracket valid?

      • Jason Brownlee April 21, 2017 at 8:33 am #

        The brackets are ignored/do nothing, and it makes the same code work in Python3.

  3. Walid Ahmed August 9, 2017 at 1:18 am #

    Hi Jason

    I could not find a models folder in my Tensorflow installation.
    Can you please help?

    • Jason Brownlee August 9, 2017 at 6:38 am #

      I think they have been removed from the most recent release.

  4. ale September 9, 2017 at 6:19 pm #

    Hi Jason

    In this page it’s written that “To make use of the GPU, only Linux is supported and it requires the Cuda Toolkit.” However, I think that Windows is also supported if Cuda toolkit and cudNN are installed. Is not it?

    • Jason Brownlee September 11, 2017 at 12:00 pm #

      It may be, it did not appear to be the case at the time of writing.

  5. Trevor Wistaff April 25, 2018 at 8:37 am #

    When getting started in machine learning would you recommend ignoring Tensorflow for now sticking to your Getting Started regime?

  6. sagar June 29, 2018 at 6:54 pm #

    Hi Jason, I have one question with respect to Tensor flow. I am trying to implement neural style transfer, using tensor flow. Was wondering if you may have any suggestions how to correct my error below. Thank you so much.


    sess = tf.InteractiveSession()[‘input’].assign(content_image))
    Traceback (most recent call last):

    File “”, line 1, in[‘input’].assign(content_image))

    File “/Users/sherrymukim/anaconda3/lib/python3.6/site-packages/tensorflow/python/client/”, line 889, in run

    File “/Users/sherrymukim/anaconda3/lib/python3.6/site-packages/tensorflow/python/client/”, line 1105, in _run
    self._graph, fetches, feed_dict_tensor, feed_handles=feed_handles)

    File “/Users/sherrymukim/anaconda3/lib/python3.6/site-packages/tensorflow/python/client/”, line 414, in __init__
    self._fetch_mapper = _FetchMapper.for_fetch(fetches)

    File “/Users/sherrymukim/anaconda3/lib/python3.6/site-packages/tensorflow/python/client/”, line 242, in for_fetch
    return _ElementFetchMapper(fetches, contraction_fn)

    File “/Users/sherrymukim/anaconda3/lib/python3.6/site-packages/tensorflow/python/client/”, line 278, in __init__
    ‘Tensor. (%s)’ % (fetch, str(e)))

    ValueError: Fetch argument cannot be interpreted as a Tensor. (Tensor Tensor(“Assign_5:0”, shape=(1, 300, 400, 3), dtype=float32_ref) is not an element of this graph.)

    • Jason Brownlee June 30, 2018 at 6:05 am #

      Sorry, I don’t have examples of tensorflow or style transfer. I cannot give you good advice.

      Perhaps post to stackoverflow?

  7. riya July 7, 2018 at 3:52 am #


    So, do I have to mention that the program shoulb be run on tensorflow CPU support or is it implicit when I do not have a GPU?

    Also, I am new to deep learning and all this API, CUDA, KERAS… etc are confusing. what are all this?


    • Jason Brownlee July 7, 2018 at 6:19 am #

      No Keras will use TensorFlow however it is configured. TensorFlow must be configured to use the GPU and I do not cover how to do that sorry.

  8. Jose November 7, 2018 at 7:54 am #

    Hi Jason!
    As always, thank you for taking the time and energy in this wonderful Website!!!
    From this section of your Linear regression example:

    I get the following error:

    NameError Traceback (most recent call last)
    28 # Fit the line.
    —> 29 for step in xrange(201):
    31 if step % 20 == 0:

    NameError: name ‘xrange’ is not defined

    • Jason Brownlee November 7, 2018 at 2:46 pm #

      Perhaps confirm that you have Python 2.7 or 3.5+ installed?

    • Carlos González February 14, 2019 at 8:16 am #

      Jose….remove “x”…just use range(201)…

  9. Khalil August 28, 2019 at 3:37 pm #

    I tried many times, but still the same issue, before that, I was using TensorFlow on the same PC. Just I have installed a window and tried to install again TensorFlow but getting an error.
    ImportError Traceback (most recent call last)
    ~\Miniconda3\lib\site-packages\tensorflow\python\ in
    —> 58 from tensorflow.python.pywrap_tensorflow_internal import *
    59 from tensorflow.python.pywrap_tensorflow_internal import __version__

    ~\Miniconda3\lib\site-packages\tensorflow\python\ in
    27 return _mod
    —> 28 _pywrap_tensorflow_internal = swig_import_helper()
    29 del swig_import_helper

    ~\Miniconda3\lib\site-packages\tensorflow\python\ in swig_import_helper()
    23 try:
    —> 24 _mod = imp.load_module(‘_pywrap_tensorflow_internal’, fp, pathname, description)
    25 finally:

    ~\Miniconda3\lib\ in load_module(name, file, filename, details)
    241 else:
    –> 242 return load_dynamic(name, filename, file)
    243 elif type_ == PKG_DIRECTORY:

    ~\Miniconda3\lib\ in load_dynamic(name, path, file)
    341 name=name, loader=loader, origin=path)
    –> 342 return _load(spec)

    ImportError: DLL load failed: A dynamic link library (DLL) initialization routine failed.

    During handling of the above exception, another exception occurred:

    ImportError Traceback (most recent call last)
    —-> 1 import tensorflow as tf

    ~\Miniconda3\lib\site-packages\tensorflow\ in
    27 # pylint: disable=g-bad-import-order
    —> 28 from tensorflow.python import pywrap_tensorflow # pylint: disable=unused-import
    29 from import module_util as _module_util

    ~\Miniconda3\lib\site-packages\tensorflow\python\ in
    47 import numpy as np
    —> 49 from tensorflow.python import pywrap_tensorflow
    51 # Protocol buffers

    ~\Miniconda3\lib\site-packages\tensorflow\python\ in
    72 for some common reasons and solutions. Include the entire stack trace
    73 above this error message when asking for help.””” % traceback.format_exc()
    —> 74 raise ImportError(msg)
    76 # pylint: enable=wildcard-import,g-import-not-at-top,unused-import,line-too-long

    ImportError: Traceback (most recent call last):
    File “C:\Users\Khalil\Miniconda3\lib\site-packages\tensorflow\python\”, line 58, in
    from tensorflow.python.pywrap_tensorflow_internal import *
    File “C:\Users\Khalil\Miniconda3\lib\site-packages\tensorflow\python\”, line 28, in
    _pywrap_tensorflow_internal = swig_import_helper()
    File “C:\Users\Khalil\Miniconda3\lib\site-packages\tensorflow\python\”, line 24, in swig_import_helper
    _mod = imp.load_module(‘_pywrap_tensorflow_internal’, fp, pathname, description)
    File “C:\Users\Khalil\Miniconda3\lib\”, line 242, in load_module
    return load_dynamic(name, filename, file)
    File “C:\Users\Khalil\Miniconda3\lib\”, line 342, in load_dynamic
    return _load(spec)
    ImportError: DLL load failed: A dynamic link library (DLL) initialization routine failed.

    Failed to load the native TensorFlow runtime.


    for some common reasons and solutions. Include the entire stack trace
    above this error message when asking for help.

  10. JG March 7, 2020 at 6:45 am #

    Hi Jason,

    I tried to follow this tensorflow tutorial to remember the “complex structure” of tensorflow works (tf 1.x version vs. Keras) and here are my main conclusions to troubles founded:

    1) Due to current tensorflow version it is 2.x, it is “eager executed”, so method s such as sessio() and run() are not directly available, your linear regression study case written for tensorflow 1.x, must be upgrade in the following ways:

    1.1) Or implementing under tf 2.x ‘eager executed’, I concluded that all your old code lines must be putted under a nest , such as :
    with tf.compat.v1.Session() as sess:

    in addition of updating some tensors with method: “tf.compat.v1”, for tensors in v2, such as:

    1.2) Or implementing tf 2.x by disabling “eager execution”, via starting your old code by the initial sentence:

    in both cases I got your same results

    1.3) Or I also tried to use directly the new methods of tf 2.x version of ‘eager training’ translating your old code such as :
    with tf.GradientTape() as tape:

    but apply to a more ‘complex structure’ such as new loss, and grad = tape.gradient(loss, w)… so I give up 🙂

    1.4) I also see that concept coming from new tensorflow 2.x is trying to apply new simple ideas of keras (wrapper such as tf.keras…) and implementing ‘eager execution’, …but now it is confusing vs. tf 1.x version …So I do not know now who is going to apply directly tensorflow under this current state …:-)


  11. JG March 8, 2020 at 3:42 am #

    OK I will do it.

    Anyway I was referring to apply pure Tensorflow to research and develop ML codes… but probably there is not much advantage, if anyone, in front of using friendly keras wrapper.

  12. Muhammad Iqbal Bazmi Hyderabad October 13, 2020 at 3:19 am #

    Upgrade it to TensorFlow 2.0

  13. Krishna November 2, 2020 at 3:47 pm #

    what is the best way to be an expert in tensorflow?

Leave a Reply