MLflow for beginners

In this tutorial, you'll learn how to manage a Machine Learning project lifecycle with MLflow, locally

Install MLflow on linux environment

pip install mlflow[extras]

NOTE: Best Practice for windows users : Work on virtual environments (pyenv or conda) within WSL2

https://github.com/pyenv/pyenv#basic-github-checkout

Explore the "titanic_classification.ipynb" notebook

This is a fairly simple classification example. It illustrates how to apply different preprocessing and feature extraction pipelines.

For more details check the link below :

https://scikit-learn.org/stable/auto_examples/compose/plot_column_transformer_mixed_types.html#column-transformer-with-mixed-types

Integrate MLflow to the pipeline :

In order to track experiments, register models and deploy them you will have to integrate some mlflow functionalities in your code, as shown in the "titanic_classification_mlflow.ipynb" notebook.

Import MLflow

    import mlflow

Set a tracking server on localhost

    mlflow.set_tracking_uri("sqlite:///mlruns.db")

Create a new experiment :

    mlflow.create_experiment(EXPERIMENT_NAME)

Set an experimment or create it if doesn't exist

    mlflow.set_experiment(EXPERIMENT_NAME)

Log runs with mlflow

Once you start a Run youu can log parameters, metrics and the model with its artifacts. Runs will be stored in mlruns directory wherever your run the code.

with mlflow.start_run(experiment_id=EXPERIMENT_ID, run_name=RUN_NAME) as run:
    
    # mlflow log parameters
    mlflow.log_param("classifier_C",param)
    
    # mlflow log metrics
    mlflow.log_metric("pipeline_test_score",accuracy_score)
    
    # mlflow log model
    mlflow.sklearn.log_model(clf, "classifier_titanic")

Launch MLflow UI on localhost with a Tracking Server:

Terminal:§ mlflow ui --backend-store-uri sqlite:///mlruns.db

MLflow UI will appear on your browser, so you can :

• Track experiments

• Track runs (informations, parameters, metrics, artifacts)

In artifacts of each run, you'll find MLmodel, conda.yaml and other files which discribes the virtual environment required to reproduce the model.

• Compare runs

• Register models

• Manage versions and transitions

You can also create registries and manipluate them, by code : https://www.mlflow.org/docs/latest/model-registry.html