search

NVIDIA-Merlin / Merlin

NVIDIA Merlin is an open source library providing end-to-end GPU-accelerated recommender systems, from feature engineering and preprocessing to training deep learning models and running inference in production.
Apache License 2.0
715 stars 111 forks source link

[QST] What is the best way of handling string UUIDs in Merlin? #1093

Open BlakeB415 opened 6 months ago

BlakeB415 commented 6 months ago

❓ Questions & Help

Details

What would be the best way of going about this?

My original idea was to either Hash or Categorify the UUID values for training, then have the user_id_raw and item_id_raw string values be returned from the Feast store during inference so that my application can get the raw UUIDs. However, Merlin does not support this paradigm as I get type errors during inference.

Another option would be to map the UUIDs to Int separately but I'm not sure how and where this should be implemented with Merlin.

Thank you in advance.

rnyak commented 5 months ago

@BlakeB415 as I get type errors during inference. please share your error, and a toy example to repro the issue. thanks.