Tensor Flow Recommenders

I’m working on a sugar tracker application for my college project. My goal is to develop a model that analyzes the types of food users prefer and then recommends foods with less sugar from those preferred food types. I’m currently focusing on building the retrieval part of the model. However, after training, the model only returns the same food name multiple times when I request three different food recommendations. Is there anyone who can help me improve the model? :")

For reference, I’ve been following this tutorial:
(Recommending movies: retrieval  |  TensorFlow Recommenders)