do I need to pass user and item matrixes to predict method for known users and items in fit step

I've fitted the model with user and items matrixes:
`model.fit( int_m, user_features = user_matrix, item_features = item_matrix)`
Let's say I want to have predictions for one user and few items, can I do like this:
`model.predict(1,[2,5,7,8,9])`
or I need to again pass matrixes like this:
`model.predict(1,[2,5,7,8],user_features = user_matrix, item_features = item_matrix))`
Im asking because results are different:
[ -2.8424215 -12.3355665 -9.75003 -8.55455 -3.7264912]
vs
[-119.41913 -117.80002 -116.015114 -117.74599 -119.52765 ]

The issue mentions model.fit and model.predict but no repository file or test. Start by checking the predict API documentation and the model.predict entry point; done means documenting whether fitted user and item features must be passed again and explaining the differing results.

Written by the indexing model from the issue text.