Use code to run an AutoML experiment

  • 3 minutes

If you want to run an AutoML experiment as part of an automated machine learning operations (ML Ops) process, you can write code to configure and initiate an AutoML experiment.

Th AutoML API provides a Python library that you can use to run AutoML experiments for classification, regression, and forecasting. To configure the specific details for an AutoML experiment, you must write code that uses the classify, regress, or forecast method as appropriate with the parameters for your specific needs.

For example, the following code runs a classification AutoML experiment.

from databricks import automl

# Get the training data
train_df = spark.sql("SELECT * FROM penguins")

# Configure and initiate the AutoML experiment
summary = automl.classify(train_df, target_col="Species",
                          primary_metric="precision", timeout_minutes=5)

# Get the best model
model_uri = summary.best_trial.model_path

Tip

For more information about using the AutoML API, see Train ML models with Azure Databricks AutoML Python API in the Azure Databricks documentation.