Collaborative Filtering - fastai vs. PySpark


I just finished up Lesson 6 where we do collaborative filtering to generate movie recommendations (tabular data). Great course thus far, I’ve learned a lot.

My question is more of a solutions design question than fastai specific per se. I have some limited experience with using PySpark for collaborative filtering. I am wondering if there are any rules of thumb as to when the fastai implementation of collaborative filtering is better suited for the problem than the PySpark version which seems more mature.

I’m not looking to start a flame war with this post, I have very limited experience with both libraries just wanted to know if anyone has any best practices to share.