[Task] Example notebooks for generation top_k recommendations with ranking and retrieval models

[rnyak]

Description

Currently none of our example notebooks do not have any code script showcasing how one can do model prediction locally, generate top-K recommendations from the prediction results. This does not include Merlin systems, it is doing offline predictions.

We can either add a new notebook or extend the existing retrieval and ranking notebooks to show such functionality. I think extending the existing notebooks would be better idea.

We have been getting requests for such examples internally or externally and this will be very informative for our users.

[rnyak]

This is still being worked on via https://github.com/NVIDIA-Merlin/models/issues/718.

I will create an example for generating topk recommendations from a ranking model, but for retrieval model we still need the https://github.com/NVIDIA-Merlin/models/issues/718 task to be completed to create an example.

Related PRs:

• https://github.com/NVIDIA-Merlin/models/pull/705
• https://github.com/NVIDIA-Merlin/models/pull/715

@sararb is currently working on TopK encoder.

[bschifferer]

@sararb do you know if the API update is finished and we can implement this example?

[rnyak]

I am closing this since we have an example for generating topK recommendations, merged by https://github.com/NVIDIA-Merlin/models/pull/900