Before we build our model, which will take a long time to train, we should create a mechanism for us to load and dump our model to the disk. If we have a way of saving the parameterization of the factored matrix, then we can reuse our model without having to train it every time we want to use it—this is a very big deal since this model will take hours to train! Luckily, Python has a built-in tool for serializing and deserializing Python objects—the pickle module.
Update the Recommender class as follows:
import pickle class Recommender(object): @classmethod def load(klass, pickle_path): """ Instantiates the class by deserializing the pickle. Note that the object returned may not be an exact match to the code in this class (if it was saved before updates). """ with open(pickle_path, 'rb') as pkl: return pickle.load(pkl) def __init__(self, udata, description...