Registering the model

The final step is to register and log the fitted model using MLflow. This allows you to track and version your models for production deployment.

The datetime, mlflow, mlforecast.flavor packages, and the fitted mlf model are preloaded for you.

This exercise is part of the course

Designing Forecasting Pipelines for Production

View Course

Exercise instructions

Set the run_name using the current timestamp created for you in the run_time variable.
Use mlflow.start_run() to start a run with the specified experiment ID.
Log the model using mlforecast.flavor.log_model().

Hands-on interactive exercise

Have a go at this exercise by completing this sample code.

# Experiment setup
experiment_name = "ml_forecast"
try:
    mlflow.create_experiment(name=experiment_name)
    meta = mlflow.get_experiment_by_name(experiment_name)
    print(f"Setting a new experiment {experiment_name}")
except:
    print(f"Experiment {experiment_name} exists, pulling the metadata")
    meta = mlflow.get_experiment_by_name(experiment_name)

# Setup the run name and time
run_time = datetime.datetime.now().strftime("%Y-%m-%d %H-%M-%S")
run_name = f"lightGBM6_{____}"

# Start the run
with mlflow.____(experiment_id=meta.experiment_id, run_name=run_name) as run:
    # Log the model
    mlforecast.flavor.____(model=mlf, artifact_path="prod_model")
    print(f"MLflow Run created - Name: {run_name}, ID: {run.info.run_id}")

Edit and Run Code