The full error will tell you which op don't have a grad impolemented for that case.
I think I responded to that elsewhere. Don't duplicate question, we have limited time to help. On Wed, Jul 20, 2016 at 5:56 PM, Aditya Gudimella < [email protected]> wrote: > I have defined a theano function which given a matrix(named array) and a > list of integers (named split_cts), splits the array into rows such that > number of rows in i th split = split_cts[i] and returns the sum of each > split joined together as a tensor. However when I try to differentiate it, > it gives me a gradient error. Specifically, the error says: > NotImplementedError: Grad is not implemented for inputs withnumber of > dimension other than 1. > > The function is as follows: > > def split_sum(array, split_cts): > ''' > Given an array like and an array of counts, it returns an array where > each element is the sum of count number of indices in array > That is retval[i] = sum(array[counts.cumsum()[i-1]:counts.cumsum()[i]]) > ''' > sep = cumsum(split_cts) - 1 > return diff(T.concatenate((T.zeros((array.shape[1], ))[None,:], cumsum > (array, axis=0)[sep]), axis=0), axis=0) > > How do I make it differentiable? Any help would be appreciated. > > -- > > --- > You received this message because you are subscribed to the Google Groups > "theano-users" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > For more options, visit https://groups.google.com/d/optout. > -- --- You received this message because you are subscribed to the Google Groups "theano-users" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/d/optout.
