Flopping for FLOPs: Leveraging equivariance for computational efficiency