Train and Deploy AutoGluon Models on Amazon SageMaker
Ever since I was introduced to machine learning, I had always being intrigued by the idea of Automated Machine Learning. I tried my hands on several libraries such as Tpot, Auto SkLearn and a few others but I never deployed any to production.
Fast forward December 2019, at re:Invent, AWS launched AWS SageMaker AutoPilot, AutoML on Amazon SageMaker that automatically trains and tunes the best machine learning models for classification or regression, based on your data while allowing to maintain full control and visibility. To get hands-on experience with AutoPilot, I attended a workshop where the AWS team demonstrated how to get started. Believe me, it was very straightforward to use. However, I am not writing about Amazon SageMaker AutoPilot.
Earlier this year, Amazon Web Services released AutoGluon, an open source library designed to enable developers to write AI-imbued apps with only a few lines of code. Ever since the release, I have attempted the available examples on Google Colab Notebooks:
Having trained the models, they were of no use if only available in notebook environment. As such, I decided to create a custom container on Amazon SageMaker to train and deploy the models. The container is only available on GitHub and works with AutoGluon Tabular Predictions — however, easily customizable for the other tasks.
Also included is automated build pipeline definition with AWS CodeBuild which rebuilds the ECR docker image once changes are pushed to the repository.
Kindly share your thoughts and comments — looking forward to your feedback. You can reach me via email, follow me on Twitter or connect with me on LinkedIn. Can’t wait to hear from you!!
