[dvduval]

Not currently possessing the skill necessary to get that library up and running, is there a way I can just test it and see how it translates?

[sixhobbits]

I guess by posting on Facebook? It's not clear from the post whether they are running this yet.

Otherwise maybe https://huggingface.co/ or similar might release a version with interface.

[Jack000]

note that the fairseq implementation requires 2x 32GB v100 gpus for inference.

[visarga]

And that's why we won't see this deployed in production anywhere.