Open topepo opened 3 weeks ago
I would make the argument here that, since the user has explicitly chosen with a non-default argument setting save_workflow = TRUE
to store that object, we ought not to warn at all, as we don't with save_pred = TRUE
I mostly agree. People might not realize that their workflow is huge (for some value of huge) but that should be relatively rare. If we keep it, we should bump the threshold up by a log unit or two.
The problem that sparked this was that simulated annealing needs the workflow and, when it calls tune_grid()
, it sets save_workflow = TRUE
even if the opposite was done in control_sim_anneal()
.
This check can be a little annoying and we have an arbitrary cutoff. This would make an argument for the cutoff.