Despite advances in deep learning, several commonly used operations in computer programming remain poorly differentiable. The pathological nature of many operations limits the set of architecture that can be trained end-to-end using gradient backpropagation. To address this limitation, Blondel et al. have developed computationally efficient differentiable sorting and ranking operators. Recently, they released an OSS PyTorch implementation of these fast differentiable sorting and ranking operators. They also demonstrate how to use Torchsort to create a differentiable Spearman’s rank coefficient function.