What causes AUTOML_EXPERIMENT_FAILED?

The target column contains too many null values or only a single class (for classification). The input DataFrame is too large for the available cluster memory. The AutoML timeout was set too short to complete even a single trial. The input schema contains unsupported column types (e.g., struct, map, array)

How do I fix AUTOML_EXPERIMENT_FAILED?

Step 1: Check the AutoML experiment UI for the error message shown on the experiment card.. Step 2: Validate the target column: ensure it has at least 2 distinct values, low null rate, and correct data type.. Step 3: Remove or cast unsupported column types (struct, map, array) before passing the DataFrame to AutoML.. Step 4: Increase the timeout parameter or use a larger cluster to give AutoML more time and memory.. Step 5: Sample the dataset if it is too large — AutoML works well with representative samples.

High severityexecution

Power BI Refresh Error:
AUTOML_EXPERIMENT_FAILED

What does this error mean?

A Databricks AutoML experiment failed before producing a best model, often because the input dataset does not meet AutoML requirements or compute resources were insufficient.

Common causes

1The target column contains too many null values or only a single class (for classification)
2The input DataFrame is too large for the available cluster memory
3The AutoML timeout was set too short to complete even a single trial
4The input schema contains unsupported column types (e.g., struct, map, array)

How to fix it

1Step 1: Check the AutoML experiment UI for the error message shown on the experiment card.
2Step 2: Validate the target column: ensure it has at least 2 distinct values, low null rate, and correct data type.
3Step 3: Remove or cast unsupported column types (struct, map, array) before passing the DataFrame to AutoML.
4Step 4: Increase the timeout parameter or use a larger cluster to give AutoML more time and memory.
5Step 5: Sample the dataset if it is too large — AutoML works well with representative samples.

Frequently asked questions

How many trials does Databricks AutoML run by default?

AutoML runs trials up to the timeout (default 1 hour for the SDK) or until it reaches diminishing returns. The number of trials varies by dataset size and problem type.

Does AutoML support time-series forecasting?

Yes — Databricks AutoML supports time-series forecasting in addition to classification and regression. Use automl.forecast() with the time column parameter.