This is the final part of the guided walkthrough.

Here, we:

  • Naviagate the commit leaderboard.
  • Make the final comments.

A lot of the decisions in an ML project are comparative by nature, and evaluating models is no different. Now, it’s time to explore the commit leaderboard!

Click on “Commit leaderboard.”

“Commit leaderboard” is on the left sidebar.

This is where we can see the full history of our project and download each version’s artifacts.

We can see that, through the tutorial, we progressed in solving the tests and ended up with a much better model in the end! The model we started with had an F1 of 0.51, while our latest model has an F1 of 0.59. In this process, however, we ended up accidentally re-introducing some issues as well, such as feature drift. This is evident from navigating between the commits and we can decide if it’s acceptable or not.

Final comments

There is still room for improvement, and that’s why you are encouraged to keep creating tests and exploring subpopulations. By following this tutorial, you have been exposed to the Openlayer method, which helps to systematically improve ML models. Furthermore, you are now capable of using Openlayer to apply the method to your own models and datasets.