In this chapter, we looked at the SavedModel serialization format. This standardized serialization format was designed with the goal of simplifying the deployment of machine learning models on many different platforms.
SavedModel is a language-agnostic, self-contained representation of the computation, and the whole TensorFlow ecosystem supports it. Deploying a trained machine learning model on embedded devices, smartphones, browsers, or using many different languages is possible thanks to the conversion tools based on the SavedModel format or the native support offered by the TensorFlow bindings for other languages.
The easiest way to deploy a model is by using Python since the TensorFlow 2.0 API has complete support for the creation, restoration, and manipulation of SavedModel objects. Moreover, the Python API offers additional features and integrations between the Keras...
