sagemaker

GNN Fraud detection on DGL

This folder contains below code,

• train RGCN model with DGL by container image in Amazon SageMaker;
• deploy the inference endpoint with code in Amazon SageMaker;
• test client for inference endpoint;
• Jupyter notebooks go through the process of training model, deploying inference endpoint and testing the inference endpoint. NOTE: make sure using the EC2(such as c5.4xlarge, m5.2xlarge or r5.xlarge), SageMaker notebook or local env with 32G+ memory and 100G+ free disk space.

Requirements of Python packages for local env,

• DGL == 0.6.*
• SageMaker >= 2.40.0 < 3.0.0
• awscli >= 1.18.140
• PyTorch >= 1.6.0 < 1.7.0
• Python >= 3.6
• pandas
• sklearn
• matplotlib