Quick-start for Merlin

Quick-start for Merlin#

This folder contains quick-start guides for Recommender System use cases. It provides scripts, powered by Merlin libraries, and best practices documentation that are designed to to allow you go quickly from a dataset of user interactions through preprocessing, training, evaluation and deployment with Triton.

We iterate in these guides over a typical Data Science process which involves data preprocessing / feature engineering, model training and evaluation and hyperparameter tuning.

Data science process with Merlin

We provide examples on the usage of those scripts with public datasets, but they are meant to be used with your own dataset without much effort.

Here are the quick-start guides currently available:

We are going to work on next releases to add:

  • Quick-start for Session-based recommendation

  • Quick-start for Retrieval

  • Quick-start for Multi-Stage RecSys