You need to scale the amount of data to take advantage of the increase in parameters. I'm not sure where we would find another 100 GitHubs worth of data.