Using Models Manager from notebooks

Models manager is automatically configured in notebook module, so you can use these capabilities without any configuration.

You can import mlflow package in any python or pyspark interpreter in order to work with the models manager

There’re two main ways to use mlflow in notebook.

Training and using models with code inside notebook

This is the most simple and easy way. You can import mlflow package and start your develop for training your model

Or you can use some model to make a prediction using a previous trained model:

Training and using models working with a git repository

In this way, notebook module works like a wrapper of some MLFlow project. MLFlow project has the code and libraries/versions definitions (using conda yaml) in the structure, so you don’t need any other configuration in order to train a model o use some trained model with the project

You can work developing your project in local enviroment, pushing your code to some git repository and having a complete control version. When you need to use the project with real data and use the computing power of the server you can get this git repo in the notebook and running it in a few lines