What is TensorFlow?

TensorFlow is a free, open-source deep learning library for Python. It makes building deep learning models easier for practitioners and researchers. There are several use cases of TensorFlow, which include models for vision, text, audio and more.

Developed by Google Brain, a deep learning artificial intelligence research team at Google, TensorFlow has a huge community and a number of resources and supporting frameworks.

Since TensorFlow’s version 2.0, Keras has become the default high level API for building and training models.

Who uses TensorFlow?

Deep learning practitioners and researchers use TensorFlow daily. Its high level APIs allow practitioners to create sophisticated state-of-the-art architectures in a few lines.

For researchers, the ability to go deep inside the framework’s machinery allows them to tweak and experiment on minute details revolving around their research.

TensorFlow comes with several useful frameworks. Tensorboard is a suit of visualization tools that allows you to monitor and understand the training of complex neural networks, which also helps in debugging. Tensorflow Lite helps create and run deep learning models on mobile and IoT devices.

The entire ecosystem of TensorFlow encompasses a wide audience who want to build and deploy deep learning models.

How can you get started with TensorFlow?

Installing TensorFlow

1. If you use pip, then type the following command to install TensorFlow as part of your Python environment:

pip install tensorflow

2. If your system has GPUs (deep learning models are computationally intensive and need GPUs to operate faster), then install tensorflow-gpu to harness the capabilities of your system. Type the following command:

pip install tensorflow-gpu

3. If you use Anaconda, then run the following command:

conda install -c conda-forge tensorflow

4. For the GPU-enabled version in Anaconda, run the following command:

conda install tensorflow-gpu

Note: Between pip and Anaconda, an Anaconda installation is preferred.

To learn more about the various functions that TensorFlow offers, you can go through the official TensorFlow documentation.

PyTorch is an alternative to TensorFlow and was built and later open-sourced by Facebook. PyTorch is generally considered to be more pythonic and intuitive. It has grown rapidly in recent times and is very popular within the machine learning community.

Keras is a high-level deep learning framework originally developed and maintained by François Chollet, a Google engineer. Initially, it was built on a TensorFlow backend and provided higher level functionalities to users. However, in 2017, Google's TensorFlow team started supporting Keras in TensorFlow's core library.

As a result, TensorFlow runs all of its higher level operations using Keras. Core TensorFlow handles only the internal and deeper functionalities.

Think we're missing something? 🧐 Help us update this article by sending us your suggestions here. 🙏

See also

Articles you might be interested in

  1. Train your first neural network: basic classification
  2. Intro to TensorFlow for deep learning
  3. Tensorflow source code