SENSEi: Input-Sensitive Compilation for Accelerating Graph Neural Networks

SENSEi optimizes GNN computations by dynamically selecting sparse-dense matrix compositions based on input attributes, achieving significant speedups.