3 ms·Scaling neural machine translation to bigger data sets with faster training3 points by stablemap 8y ago