Open sgaseretto opened 4 years ago
@sgaseretto Have you found an answer ? I'm facing the same issue
Within predict, you pass in the item ids that you want included to get predictions for. I would suggest creating a function that returns only item ids that are still in the system, and using this to filter which items you get prediction for. For example:
predictions = predict(user_ids=1, item_ids=items_in_system_right_now())
How can this be done with LightFM? Imagine you have a very good model already trained, and you discontinue a large group of products, and also some of your users deleted their accounts. How can this user and items representations be deleted from the model in order to avoid having to train a new model without them?