Boston Housing In-Depth

Now we will look at deploying a model using the low level approach. This method requires us to describe the various properties that our endpoint should have and what inference code and model should be used.

To follow along, open up the Boston Housing - XGBoost (Deploy) - Low Level.ipynb notebook in the Tutorials folder.

Using the low level approach to deploy our model requires us to create an endpoint, which will be used to send data to our model and to get inference results.

In order to create an endpoint in SageMaker, we first need to describe an endpoint configuration. This describes to SageMaker the various properties we want our endpoint to have. Once we've created the endpoint configuration we can ask SageMaker to create an endpoint with the properties we want.

The actual endpoint that is created by SageMaker is a combination of a compute instance (some remote server) running a docker container with the inference code on it and a URL that data can be sent to and returned from. This URL is used as an interface to the compute instance, which receives data, performs inference using our model and returns the result.