[Question] Did you use Discriminator for recent results

[MaksymDel]

Hi,

It is clear that you used discriminator for the models in the paper Unsupervised Machine Translation With Monolingual Data Only.

But do you also use it to obtain results for NMT models presented in Phrase-Based & Neural Unsupervised Machine Translation paper?

From the code it seems like if --n_dis parameter equals 0 (which is default), the discriminator is not used. You found it to have no positive effect?

Thanks!
Maksym

[glample]

Hi,

Indeed, in the last paper we didn't find any benefits using the discriminator. That being said, we didn't tune it very well, but I don't think it would make a big difference even if well tuned.