Very slow get_learner for collaborative filtering

Part 1 (2018)
1

When doing collaborative filtering for the movie-lens dataset, calling cf.get_learner always takes a very long time. Does anyone know if this is the expected behavior for this function? I know the image learners did a decent amount of data processing for pre-computed activations and their data function also resized images, but this should just be reading from a csv for everything.

Selection_023
Selection_023.png833×428 35.7 KB

Thanks for any help!

EDIT: RESOLVED. I updated everything and it works now. It only takes 1.36s now.

2

that does seem very long. I ran the notebook as is and my wall time was 2.73s for that step. In fact the entire notebook was able to run and finish before I could finish this reply.

1 Like
3

RESOLVED. I updated everything and it works now. It only takes 1.36s now.