When we process a WSI/part of WSI, we have access to some extra information that's not being used by the model - the level of magnification.
In this task, we create a method of conditioning the model on the resolution provided: perhaps starting with something like AdaLN.