Hey! Sorry, didn't fully read your question and missed that you already did it. It should not be done inside the clearm-serving-triton service but instead inside the clearml-serving-inference service. This is where the preprocessing script is ran and it seems to be where the error is coming from.