Gradient Weights for Softmax Function

Hi.  In the paper you derived equations for the case when softmax function is used to obtain final class probabilities. (Equations 21-24) 
I am unable to find an implementation of these formulae in this repository. Any help is greatly appreciated.