New Features
- Serverless and Serving Endpoint compatibility
- NVIDIA's MONAI framework with Databricks compatibility layer
- Incremental processing with spark structured streaming and autoloader
- Automatic zip extraction
- Compatibility with UC Volumes
- OHIF Viewer with Databricks compatibility layer
MONAI integrations
- Model wrapper for MONAILabel to serve as PyFunc ML Model usable in Serving Endpoints
- MONAILabelTransformer to call MONAILabel server in a serving endpoint
- MONAILabel server instructions to notebook 05-Monailabel.py
- A new example to create a custom model with user-defined labels in notebook 05-Monailabel.py
- New workflow example to use pixels with serving endpoints
- MONAILabel radiology app and model creation under monailabel_model folder
- MONAILabel in serving endpoint connectivity to the lakehouse app
Others
- Added SQL statements to create Pixels' tables with optimal configuration during the initialization of the Catalog class
- Added deep parameter in DicomMetaExtractor for optimization purposes, set it as false to avoid scanning dicom pixel data during metadata extraction
- Fixed governance notebook sample
- Fixed RUNME
- Code cleanup using check style
- Removed unnecessary notebooks
Full Changelog: 0.0.6...2.0.0