I was wondering if that impacts the accuracy, obviously the rtk output wasn't in the training dataset, but maybe it doesn't matter at the end