It would be great if we can run evaluators without inference or weight learning. Maybe if `--eval` is supplied without `--infer` or `--learn`.