Ability to use a Runtime Parameter in the Model’s dependent feature or a Dataset’s target feature.
Recording Training and Validation Data for Models using Datasets registered within the system in addition to the existing method of documenting an external location.
Simplification of some properties for Models in notebooks. Exposing details such as alias, colname, and type of the model to be consistent with other objects in the platform.
This has 2 breaking changes
model.type -> model.algorithm_type (see below)
Model(name=)-> Model(alias=) (see below)
Improvements in Table Lookup models to handle corner cases and boundary conditions in a more flexible way.
Common
Custom File upload Fields
Admins can now create Fields in Settings, which can be added on the Registry pages for various objects.
These fields behave as custom attributes for objects like DataElement, Feature, Dataset, Model, and Policy to which file attachments can be added.
Job Execution enhancements
Ability to tag jobs with Keywords while running a job on the platform.
Users can now set the start time to specify when the job should run in the future.
Allow selection of multiple objects as challengers in comparison and multiple simulations as benchmarks in a validation.
Support for job configurations which can be used to configure the execution environment before running a job. This can be used to configure the environment where the job is running (like the spark session with custom configs).
Show the time breakup for the Job - which gives a breakup of the time taken to execute various parts of the job.
Change Management
A new tab “Impact assessment” which can be used to understand what the impact of creating new version of an object can be downstream - it gives a dashboard of how many dependent objects are using the old version and their update status.
Change history will now show alerts for changes that have occurred that the user has not seen. Users can mark these changes as read/unread in the Change History.
Approval Workflow:
Ability to choose reviewers based on their role by selecting the "Specific Roles" option when defining an Approval Responsibility.
Allow Runtime Parameters to be Array types (like ArrayNumerical, ArrayCategorical, ArrayDateTime).
Technical
Multiple performance improvements for various APIs that are Improve speed of multiple APIs.