Highlights Keywords Environment, Climate/Weather, Engineering Challenge Graph Neural Networks (GNNs) are a key technology for learning from graph-structured data and are expected to drive future AI applications on large-scale high-performance computing (HPC) systems. These models require long sequences of sparse-dense matrix multiplications, particularly involving binary adjacency matrices and dense node feature embeddings. Existing sparse matrix formats like COO and CSR are designed to store only non-zero entries, but they do not exploit similarities between matrix … Continue reading Success Story: CBM4scale