Training and Serving Merlin on AWS SageMaker

With AWS Sagemaker, you can package your own models that you can train and deploy in the SageMaker environment. The following notebook shows you how to use NVIDIA Merlin for training and inference in the SageMaker environment.

• Training and Serving Merlin on AWS SageMaker

This notebook assumes that readers are familiar with some basic concepts in NVIDIA Merlin, such as:

• Using NVTabular to GPU-accelerate preprocessing and feature engineering.

• Training a ranking model using Merlin Models.

• Making Inference with the Triton Inference Server and Merlin Models for Tensorflow.

To learn more about these concepts in NVIDIA Merlin, see for example Deploying a Multi-Stage Recommender System in this repository or example notebooks in Merlin Models.

Running the Example Notebook

You can run the example notebook with the latest stable merlin-tensorflow container. See Running the Example Notebooks for more details.

Additionally, you need to configure basic AWS settings. For setting up AWS credentials, refer to Configuring the AWS CLI in the AWS documentation. After you configure basic AWS settings, you can mount your AWS credentials by adding -v $HOME/.aws:/root/.aws to your Docker command in Step 1 of Running the Example Notebooks:

docker run --gpus all --rm -it \
  -p 8888:8888 -p 8797:8787 -p 8796:8786 --ipc=host \
  -v $HOME/.aws:/root/.aws \
  <docker container> /bin/bash

and follow the remaining steps in Running the Example Notebooks.