GPUs are specialized hardware components that were originally developed for graphics applications. However, it was realized that GPUs could be used to perform computations in a DNN architecture.

In the second part of this chapter, we saw how to install a GPU-enabled TensorFlow and how to manage GPU devices.

In the next chapter, we will discuss some advanced programming features using TensorFlow by integrating with other third-party tools such as, Keras, PrettyTensor and TFLearn. In addition, we’ll give an overview of some TensorFlow-based framework--Keras, PrettyTensor and TFLearn.

We'll show the main topics of these frameworks, with very interesting application examples.