Training a recommender system on a large curated dataset

In the Training a recommender system on a small curated dataset section, we saw the basics of how to create a recommender system model. The resulting system left something to be desired because the dataset only included user IDs and movie IDs, so it wasn't possible to determine what movies were actually being rated by users and having their ratings predicted by the model.

In this section, we are going to create a recommender system that addresses this gap in the previous recommender system because it is trained on a dataset that includes movie titles. Like the ML_SAMPLE dataset, the dataset we'll use in this section, ML_100k, is also derived from the MovieLens dataset, but it includes a much larger set of records and a much richer set of features. By creating a recommender system using this dataset, we will encounter additional features in fastai for ingesting and working with recommender system datasets and get...