Hacker News new | ask | show | jobs
by lamontcg 926 days ago
Practically does using unified memory mean that the slow transfer of training/testing data to the GPU would be eliminated?