Losing Heads in the Lottery: Pruning Transformer Attention in Neural Machine Translation

30676

Knowledge Triples