What have we learned so far?

In this lesson we took a look at how we can use SageMaker to tune a model. This can be helpful once we've found a good base model and we want to try and iterate a bit to refine our model and get a little more out of it.

We also looked at using CloudWatch to monitor our training jobs so that we can better diagnose errors when they arise. This can be especially helpful when training more complicated models such as those in which you can incorporate your own code.

What's next?

In the next lesson we will take a look at updating a deployed model. Once you've developed a model and deployed it the story is rarely over. What happens if some of the built in assumptions about your model change over time?

We will look at how you can create a new model which more accurately reflects the current state of your problem and then update an existing endpoint so that it uses your new model rather than the original one. In addition, using SageMaker, we can do this without needing to shut down the endpoint. This means that whatever application is using your deployed model won't notice any sort of disruption in service.