Skip to content

On-premise users: click in-app to access the full platform documentation for your version of DataRobot.

Feature cache for Feature Discovery deployments

Availability information

Feature Cache for deployed Feature Discovery models is off by default. Contact your DataRobot representative or administrator for information on enabling the feature.

Feature flag: Enable Feature Cache for Feature Discovery

Now available as public preview, you can schedule feature cache for Feature Discovery deployments, which instructs DataRobot to pre-compute and store features before making predictions. Currently, you can only make batch predictions with Feature Discovery projects; however, generating these features in advance makes single-record, low-latency scoring possible.

Once feature cache is enabled and configured in the deployment's settings, DataRobot caches features and stores them in a database. When new predictions are made, the primary dataset is sent to the prediction endpoint, which enriches the data from the cache and returns the prediction response. The feature cache is then periodically updated based on the specified schedule.

To enable feature cache, go to the Predictions > Settings tab of a Feature Discovery project's deployment. Then, turn on the Enable Feature Cache toggle and choose a schedule for DataRobot to update cached features.

Note

If you are configuring the settings for a new deployment, the creation process may take longer than usual as features are computed and stored for the first time during deployment creation. Once feature cache is enabled for a deployment, it cannot be disabled later on.

You can change how often DataRobot caches features or monitor the status of feature caching on the deployment's Predictions > Settings tab.

General statuses

In your deployment's settings, you can monitor the status of feature cache.

The table below describes each possible status:

Status Description
Not fetched Feature cache was configured but data hasn't been populated into feature cache yet. Predictions are impossible at the moment.
Outdated Data was not populated during the last scheduled run. Outdated data still present in feature cache and predictions are possible but accuracy can be badly impacted.
Configuration failed Feature cache was enabled but failed to be configured. Predictions are impossible.
Failed to fetch Data failed to be stored in cache. Predictions are impossible.
Updated Last scheduled run was completed successfully. Predictions work as expected.

Considerations

Consider the following when enabling feature cache for a Feature Discovery project:

  • The maximum number of prediction features is 300.
  • The scoring dataset can have a maximum of 200 rows. Datasets with more than 200 rows will not be scored with feature cache.
  • Feature cache is not compatible with data drift tracking.
  • Feature cache is only visible in the UI if DataRobot detects secondary datasets.
  • Feature cache cannot be disabled once it is enabled for a deployment.

Updated April 11, 2023