Clem rev3

Created by: ClemSc

Hey, it would be good to test my changes by running an evaluation on the test set with any model and make sure that the code go without error. I tested what I could, but on this server I cannot run the whole inference. If it works, its good to merge on release