Ebonite CoreΒΆ
This module is responsible for model analysis and model persisting. If you use vanilla ebonite, this is mainly what you are working with.
Main model analysis API abstractions are
DatasetHook
- hook for dataset analysisDatasetType
- dataset descriptorModelHook
- hook for model understandingModelWrapper
- model wrapper for different ml librariesModelIO
- model input / output for different ml model serializers
Main model persisting abstractions are
MetadataRepository
- Repository to store model metadata (like sql database)ArtifactRepository
- Repository to store model artifacts (like s3 or nexus)Ebonite
- Main user-facing ebonite client class
Also these helper functions available:
create_model()
- createsModel
instance from model object and sample data