Hands-on MLFlow: managing the end-to-end machine learning lifecycle in practice
Organized by SIT Academy
MLFlow is one of the most popular tools to manage the machine learning lifecycle from the beginning to the end. In this one-day workshop, we propose to guide you through the various ways you can use this platform to track, package, deploy and share your models.
As the pressure increases to bring models to production as fast and reliably as possible, the practices of machine learning operations (MLOps) are becoming the standard for data-driven companies. However, the abundance of tools to manage each aspect of the machine learning pipeline can appear daunting.
In this workshop, we focus on MLFlow, an open-source platform that proposes solutions for the four primary pain points when it comes to managing the full machine learning life cycle:
- tracking metadata on your running experiments,
- packaging data science code in a reusable way,
- managing and deploying models
- model versioning and sharing
We propose a day-long, hands-on tutorial to introduce MLFlow tools through examples and exercises.
Data scientists, statisticians, and AI enthusiasts who have developed machine learning models using Python and want to gain practical experience in managing the machine learning lifecycle, end-to-end. A preparation document will be provided a few weeks before the workshop.
Time Frame: June 22, 9:00 – 12:15, 13:15 – 16:30
- Introduction: The Machine Learning Lifecycle and how MLFlow can help
- Sequence 1 (9:00 – 10:30): MLFlow Tracking: log parameters, code versions, metrics, and artifacts when running your machine learning code.
- Sequence 2 (10:45 – 12:15): MLFlow Projects: package data science code in a reusable way
- Sequence 3 (13:15 – 14:45): MLFlow Models: package machine learning models and
- Sequence 4 (15:00 – 16:30): MLFlow Model Registry: get familiar with the centralized model store
- Summary of the day and further resources.
Max. participants: 20