TensorFlow

TensorFlow was initially developed by the Google Brain team to be used internally at Google, and was then to be released to the larger public. On November 9, 2015, it was distributed under the Apache 2.0 open source license, and since then it has become the most widespread open source software library for high-performance numerical computation (mostly used for deep learning). It is capable of computations across a variety of platforms (systems with multiple CPUs, GPUs, and TPUs), and from desktops to clusters of servers to mobile and edge devices.

In this book, we will use TensorFlow as the backend of Keras, that is, we won't use it directly, but we will need to have it running on our system:

• Website: https://tensorflow.org/
• Version at the time of print: 1.8.0

Installing TensorFlow on a CPU system is quite straightforward: just use pip install tensorflow. But if you have an NVIDIA GPU (you actually need a GPU card with CUDA Compute Capability 3.0 or higher) on your system, the requirements ramp up and you first have to install the following:

• CUDA Toolkit 9.0
• The NVIDIA drivers associated with CUDA Toolkit 9.0
• cuDNN v7.0

For each operation, you need to accomplish various steps depending on your system, as detailed on the NVIDIA website. You can find all the directions for installation depending on your system (Ubuntu, Windows, or macOS) at https://www.tensorflow.org/install/.

After having accomplished all the necessary steps, pip install tensorflow-gpu will install the TensorFlow package that's optimized for GPU computations.