Binarized Neural Machine Translation

Yichi Zhang, Ankush Garg, Yuan Cao +4

The rapid scaling of language models is motivating research using low-bitwidth quantization. In this work, we propose a novel binarization technique for Transformers applied to mac…