Fully Quantized Transformer for Machine Translation
This addresses the computational efficiency problem for machine translation practitioners, offering a novel solution that achieves state-of-the-art quantization results.
The paper tackles the problem of high computational costs in neural machine translation by proposing FullyQT, a fully quantized Transformer, and shows that it avoids any loss in translation quality, with 8-bit models scoring greater or equal BLEU on most tasks compared to full-precision models.
State-of-the-art neural machine translation methods employ massive amounts of parameters. Drastically reducing computational costs of such methods without affecting performance has been up to this point unsuccessful. To this end, we propose FullyQT: an all-inclusive quantization strategy for the Transformer. To the best of our knowledge, we are the first to show that it is possible to avoid any loss in translation quality with a fully quantized Transformer. Indeed, compared to full-precision, our 8-bit models score greater or equal BLEU on most tasks. Comparing ourselves to all previously proposed methods, we achieve state-of-the-art quantization results.