yes, that's one trick they did in the paper. they also do pruning, weights close to zero are removed. then they implemented a gpu sparse matrix multiplication

yes, that's one trick they did in the paper. they also do pruning, weights close to zero are removed. then they implemented a gpu sparse matrix multiplication kernel.