creating a dataset that loads the multi-modal image and randomly selects one modality: this is the YuccaTrainDataset_1modality
adding support in the manager for overriding configs. This is by making the staticmethod that subclasses can run using super and then override as needed. Possibly something that should be extended to other parts, but this is not so relevant for Agave.
MANUALLY overriding the "num_modalities" key in the plans and the num_classes in the PlanConfig. Regardless of what the data looked like, and is registered to look like in the plan, we want the input channels to be 1 (controlled by num_modalities) and the output channels to be 1 (controlled by n_classes)
@asbjrnmunk
essentially involves