DeepGNN, A Framework For Training Machine Learning Models On Large Scale Graph Data

DeepGNN is framework used internally at LinkedIn and Microsoft for training ML models on large graphs. The biggest challenge to our data scientists is to scale up GNNs to massive graphs with billions of nodes and edges. To address this challenge, we developed DeepGNN framework. It allows training models on large datasets by serving the graph in a distributed fashion with graph engine servers. In this talk, we will highlight design and strengths of DeepGNN, such as efficient memory layout, sampling, and support for both PyTorch and TensorFlow