.svg)
.svg){kind=small}
Hopsworks 3.8 is now generally available. This new release adds a notification service to the Hopsworks platform allowing changes to Feature Groups to result in notifications being sent to relevant subscribers. In addition a number of improvements have been made to transformation functions including model dependent transformations and on-demand transformations. In addition a number of minor improvements were included with this release.
Notification Service
With the release of 3.8, Hopsworks now provides a notification service that allows developers to be notified when changes occur to Feature Groups enabling feature engineering to be improved when data changes that match specific criteria.
Model Dependent Transformations
Model-dependent transformations in feature engineering refer to the process of modifying or creating new features based on the specific requirements or assumptions of a machine learning model. Some models, like linear regression or logistic regression, assume linear relationships between features and the target. Hopsworks now supports both model dependent and independent transformation functions.
On-Demand Transformations
On-demand transformations in feature engineering refer to the dynamic generation of features as needed when making predictions. Unlike pre-processed features that are fixed before inference, on-demand transformations are applied in real-time, often in response to changing data or specific model requirements. With the release of 3.8, Hopsworks now support On-Demand Transformation functions.
Bug Fixes
Featurestore
FSTORE-1406 - Delete Hive table if it exists when deleting feature group
FSTORE-1405 - Add detailed status field to RonDB Rest Server
FSTORE-1415 - Add helper column support for Online Store REST Client
Ancillary Services
HWORKS-1235 - Ensure model deployment consistency between Hopsworks and Kubernetes
HWORKS-1197 - Add support for case insensitive search on datasets api
HWORKS-1296 - aircompressor security advisory