End-to-end session-based recommendation

These end-to-end example notebooks focus on the following:

  • Preprocessing the Yoochoose e-commerce dataset.

  • Generating session features with on GPU.

  • Using the NVTabular dataloader with 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.

Refer to the following notebooks: