Scaling to Large Datasets with Criteo

Criteo provides the largest publicly available dataset for recommender systems. The dataset is 1 TB uncompressed click logs of 4 billion examples. The example notebooks show how to scale NVTabular in the following ways:

• Using multiple GPUs and multiple nodes with NVTabular for ETL.

• Training recommender system model with NVTabular dataloader for PyTorch.

Refer to the following notebooks:

• Download and Convert

• ETL with NVTabular

• Training a model: HugeCTR | TensorFlow | FastAI

• Use Triton Inference Server to serve a model: HugeCTR | TensorFlow