Hands-On Industrial Internet of Things

Understanding the implementation of a SageMaker container

As mentioned previously, SageMaker supports a standard container, but also allows us to build our own algorithms with a custom container. In this exercise, we are going to build a container with our own dependencies. These include Keras, Pandas, and NumPy, but also NGINX (a popular web server to build microservices: https://www.nginx.com/) and Flask (the Python support to web servers) to expose the model through a REST API. SageMaker supports two execution modes, training, where the algorithm uses input data to train a new model, and serving, where the algorithm accepts HTTP requests and uses the previously trained model to make an inference, which is called a prediction.

The following ...