Introduction
Up to this point, we have covered how to train and evaluate various models in TensorFlow. In this chapter, we will show how to write code that is ready for production usage. There are various definitions of production-ready code but, for us, production code will be defined as code that has unit tests, separates the training and evaluation code, and efficiently saves and loads various needed parts of the data pipeline and created graph session.
Note
The Python scripts provided in this chapter should be run from the command line. This allows tests to be run, and device placements to be logged to the screen.