End-to-end session-based recommendation

This end-to-end example notebook is focuses on:

• Preprocessing the Yoochoose e-commerce dataset

• Generating session features with on GPU

• Using the NVTabular dataloader with the Pytorch

• Training a session-based recommendation model with a Transformer architecture (XLNET)

• Exporting the preprocessing workflow and trained model to Triton Inference Server (TIS)

• Sending request to TIS and generating next-item predictions for each session

Notebooks

• End-to-end session-based with Yoochoose