I am really not sure whether this belongs here or in butcher/workflows/recipes. Happy to move this issue accordingly. Anyway, when preparing a model, one really important consideration is making sure any training data is removed. It is my understanding that vetiver calls butcher under the hood to trim down the workflow object, but the butcher methods for recipe objects don't seem to strip the training data.
Some reading into ?recipes::prep suggests that if you use retain = FALSE then this data won't be included, but prep() is generally called internally during training and so I'm not sure how to avoid bringing the data along for the ride.
Am I using this wrong? Do I need to remove this data manually? I am a bit confused at the best way to ensure training data is stripped from a model/workflow before deployment.
Here is a reprex showing the training data is embedded in the vetiver model.
Seems like maybe I just needed to update some packages...will close this but leave it as a relic for other to remember to update your package before filing issues 😬
Hi,
I am really not sure whether this belongs here or in
butcher
/workflows
/recipes
. Happy to move this issue accordingly. Anyway, when preparing a model, one really important consideration is making sure any training data is removed. It is my understanding thatvetiver
callsbutcher
under the hood to trim down theworkflow
object, but thebutcher
methods forrecipe
objects don't seem to strip the training data.Some reading into
?recipes::prep
suggests that if you useretain = FALSE
then this data won't be included, butprep()
is generally called internally during training and so I'm not sure how to avoid bringing the data along for the ride.Am I using this wrong? Do I need to remove this data manually? I am a bit confused at the best way to ensure training data is stripped from a model/workflow before deployment.
Here is a reprex showing the training data is embedded in the vetiver model.