Types processing system implemented
Created by: Dreamlone
A system for handling types in tabular data has been implemented for the framework. Now, if a dataset contains columns with several data types (e.g. str and float), this column will be converted to a suitable (single) format or deleted. The same system of type processing is applied to target columns. For classification task if there is a conflict in the target column, all values will be cast to string type.
The system for determining the presence of categorical data in tables has also been modified. Now all column types are stored in a separate specialized field SupplementaryData - column_types. All information about trait types and target types is stored in this field throughout the training in the pipeline