Tobias Sterbak: Introduction to MLOps with MLflow

Discover the basics of MLOps and the MLflow tool, learn how to track, manage, and reproduce machine learning models from development to deployment and maintenance.

Key takeaways

MLOps (Machine Learning Operations) involves managing the lifecycle of machine learning models, from development to deployment and maintenance.
MLOps includes tracking, versioning, and reproducibility of models, as well as managing model dependencies and using a standardized format for packaging models.
MLflow is a tool for MLOps that provides a simple, Python-based API for tracking and managing machine learning models.
MLflow includes three main components: tracking, where it collects information about the training and validation of models; models, which is a specification of how models can be packaged and reproduced; and artifacts, which is a way to manage and store model artifacts.
In MLflow, you can track and manage the lifecycle of your models, including the ability to version your models and track their dependencies.
MLflow also includes a client API that allows you to access and interact with the tracking server and the model registry.
The goal of MLOps is to make it easier to manage and maintain machine learning models, and to provide a way to reproduce and track the results of your models over time.
MLflow is designed to be used in conjunction with other tools and platforms, such as Databricks and Azure ML, to provide a complete end-to-end solution for MLOps.
In the talk, the speaker demonstrated how to use MLflow to track and manage the lifecycle of a machine learning model, including how to track metrics and parameters, and how to version and reproduce the model.

Tobias Sterbak: Introduction to MLOps with MLflow

More talks