- Document Understanding Release Notes
- ML Packages and Public Endpoints Release Notes
Document Understanding Release Notes
June 2021
Auto-retraining of ML Models
Auto-retraining is now available in Public Preview. Now, you can import data from Validation Station and use it to retrain ML Models without manual steps in the process.
The main benefits are ease of use and the long-term maintainability of the ML Models. You can build your ML Model, label data, and then set up this automatic retraining and the ML Model will keep itself up to date.
Usability improvements
Implemented dark theme support.
Export validation
Export validation errors are reported much sooner.
Test sets no longer require export validation.
Fast, Robust, and Scalable Experience
It is with great excitement to announce that Cloud Data Manager is now generally available. From this point forward, you can manage your Document Understanding datasets at the convenience of Automation Cloud. Highlights include:
- Faster startup.
- Faster labeling.
- No limit to volume of documents that can be imported or exported.
- Friendlier Export dialog.
- Limit of 1500 documents per import, though successive imports allow you to import as many documents as you need.
Migrating Data Manager sessions
Cloud Data Manager has been in Public Preview since May 5, 2021. All Data Labeling sessions created before May 5, 2021 will be deprecated on July 1, 2021, and all labeled data in these sessions need to be migrated to new sessions by July 1, 2021.
Old sessions may be identified by their Status in the Data Labeling view of AI Center.
The image below shows one new session with the status Available and one old session with the status Stopped (MIGRATE BY July 1).