Hi, By default the code splits the data from each collaborator 80/20 for local training and local validation. To speed up training, I'd like the option to completely disable local validation while trying out ideas. Is that something easily implemented with the existing infrastructure?