Did you ever find a solution to this? I’m having a similar issue with my coreml model eating up too much memory when predicting.