Hello I was wondering what was the best way to implement col-major convolutions in Cudnn (aside from just reformatting all the data).
My original project utilized cuBLAS so my tensors are set in col-major order. It seems strange cudNN is written for row-major tensors as this makes integrating the two libraries much more difficult.
Is there a simple-hack that I’m missing to integrate the two together?
-Thanks again, JJ