Comment by egyptianblue

Is there a generalizable approach that you have found for finding ways to be more efficient with the models, or is it case-by-case?