Researchers from UW-Madison have OSS’ed Marius, a system for large-scale graph embeddings. Most existing approaches to building graph embedding models require model developers to leverage expensive distributed computing tools and platforms. In contrast, Marius enables users to learn graph embedding models over billion-edge graphs on a single machine. Marius, which offers a configuration-driven programming model, removes the data movement bottleneck by leveraging a new data replacement policy to maximize GPU utilization and exploit the full memory hierarchy.