Dear Organizers, Under the rules, there is a statement, which says: "The containerized solution must be developed in a way where the method is trained using the training images available on the Cheaha system. The containerized solution cannot be a fully trained model that only makes predictions on the test images." Until now I think many of the competitors (we surely did) submitted locally trained models and the Docker only made the predictions to make life easier. However, as we are getting closer to the final deadline, we would like to meet all the rules. Can the organizers please confirm if it is still necessary to train all the models within the docker to have a valid submission at the end, or it is only a legacy rule from the time when only 50 training data was supposed to be released? Thank you, Balint