3 ms·Shifting deep learning research from SOTA accuracy to SOTA efficiency1 points by austin_kodra 7y ago