@sgugger I’m looking for a built-in method to get all possible movie ratings per user (complete all the NaNs in the cross-tabular view of users vs movies). As an end goal, I’d like to be able to pass any user and movie ID to get a prediction as in predict(userId, movieId)
.
When I call learn.get_preds(ds_type=DatasetType.Test)
it only returns predictions on the existing movie-user ratings. The workaround I use is to pass all possible user-movie combinations as a test set:
data = CollabDataBunch.from_df(ratings, seed=42, pct_val=0.1, user_name=visitor, item_name=content, rating_name=score, test=ratings_all)
What I do feels ‘hacky’ so I wonder if there’s something built-in that I missed. Is there a method to get predictions on all possible user-movie ratings?
Thank you!