Getting Started with Merlin and the MovieLens Dataset

The MovieLens25M is a popular dataset for recommender systems and is used in academic publications. Most users are familiar with the dataset and we will teach the basic concepts of Merlin:

• Learn to use NVTabular for using GPU-accelerated feature engineering and data preprocessing.

• Become familiar with the high-level API for NVTabular.

• Use single-hot/multi-hot categorical input features with NVTabular.

• Train a Merlin Model with Tensorflow.

• Use the Merlin Dataloader with PyTorch.

• Train a HugeCTR model.

• Serve recommendations from the Tensorflow model with the Triton Inference Server.

Explore the following notebooks:

• Download and Convert

• Feature Engineering with NVTabular

• Training with TensorFlow

• Training with PyTorch

• Training with HugeCTR

• Serve Recommendations with Triton Inference Server (Tensorflow)