What have we learned so far?

In this lesson we learned how to deploy a model that has been created using SageMaker. We took a look at how to construct endpoints and how to use those endpoints to send data to a deployed model.

In addition, we looked at what we needed to do if we wanted anyone to have access to our deployed model. To make this work we first implemented a Lambda function that took care of data processing and interacting with the model. Then we created an interface through which we could send data to our Lambda function using API Gateway.

How does the data flow?

QUESTION:

Describe how the data flows through our sentiment analysis web app.

ANSWER:

To begin with, the user enters a review on our website.

Next, our website sends that data off to an endpoint, created using API Gateway.

Our endpoint acts as an interface to our Lambda function so our user data gets sent to the Lambda function.

Our Lambda function processes the user data and sends it off to the deployed model's endpoint.

The deployed model perform inference on the processed data and returns the inference results to the Lambda function.

The Lambda function returns the results to the original caller using the endpoint constructed using API Gateway.

Lastly, the website receives the inference results and displays those results to the user.

What's next?

In the next lesson we are going to look at hyperparameter tuning. This is a method by which we can test a variety of different hyperparameters and chose the ones that work best for our data set.

While doing this we will also take a look at CloudWatch, which is a service that allows us to look at the logs generated by the various SageMaker tasks we perform.