AutoRec: Rating Prediction with Autoencoders

https://d2l.ai/chapter_recommender-systems/autorec.html

How can the training loss be so low?

Thank you for this awesome book!