About the parameter "row" in your experiment for imagenet

Hi, 
I noticed the use of the "row" parameter in your code, which is present in all experiments related to fixing convolutional neural networks. However, I couldn't find its meaning in your paper and code. Could you please explain what this parameter is used for ?