3 ms·Speeding Up Transformer Training and Inference by Increasing Model Size2 points by jonbaer 7y ago