mml.core.data_loading

mml.core.data_loading deals with all file and data operations except for the initial data integration part, which is dealt with in mml.core.data_preparation.

Data loading from a task is layered in three:

• TaskStruct holds only lightweight meta information of a task

• TaskDataset implements mechanics to use the task description file and load individual samples

• MultiTaskDataModule holds different splits of (potentially) multiple task, deals with transforms and provides the dataloaders

Furthermore file_manager holds the central MMLFileManager. augmentations deals with data augmentations. task_attributes declares task meta information formats as they are needed by the central task_description. modality_loaders implement the reading operations for different Modality’s. Lastly utils holds a small number of utility functions for data transformations.