Creating recommendations for single users without leveraging user identity features

Nobody has claimed this yet.

Hello,

I am creating recommendations for a single user with a model that was trained with user features but without user identity features (i.e. have set `user_identity_features = False` when creating the Dataset class). However, I encountered a shape issue when the method `_construct_feature_matrices` is called in the `predict` method. I noticed that the mismatch in shape was caused by this line (Line 851) in the `predict` method:
`n_users = user_ids.max() + 1`

From what I can understand, user_ids is an array as long as the number of items but every element is the index value of a user (if a single user id is provided). However, shouldn't it be the number of users if user_features is provided?

Thanks ahead.

Start in the predict method at line 851 and trace how user_ids and user_features are passed to _construct_feature_matrices. Reproduce prediction for a single user with user_identity_features=False, then inspect the resulting shapes. Done means the prediction path accepts the provided user features without a shape mismatch.

Written by the indexing model from the issue text.