Transformer big

[shaochenze]

It's an interesting work! I'm trying to reproduce your result. Have you tried to implement your method in Transformer big? What result can you obtain in WMT14, and could you provide the training configurations?

[SunbowLiu]

We only evaluated our proposed method on base Transformer. As the method is mostly architecture-agnostic, I think most of the hyperparameters can also work well on big Transformer.