Summary

In this chapter, you learned the advanced techniques required to train models at scale using different distribution strategies. You further reviewed best practices for hyperparameter tuning to find the best version of the model to meet your objectives. You learned how to organize and track multiple experiments conducted in a typical ML workflow and create comparison reports.

Using the SageMaker capabilities and best practices discussed in this chapter, you can tackle ML at scale, allowing your organization to move out of the experimentation phase. You can take advantage of large datasets collected over years, and move toward realizing the full benefits of ML. In the next chapter, you will continue to enhance ML training by profiling training jobs using Amazon SageMaker Debugger.