How to minimize an optimizer in a multigpu distribution

Question

I\'m trying to adapt a DL model of MNIST to be run in many GPUs at once. However, I can\'t find a way to make it work. I am new to DL so I don\'t fully understand all the lo