Julia Kreutzer

Yes, very good points. My concern was mostly about the translate mode, where we don't have targets, and also no sacrebleu computation. We don't want to filter, but we want...

Hi @chrisemezue! Thanks for the code! I think it needs some refactoring before we can merge it, in order to make it easy to read and use and compatible with...

Hi @chrisemezue, thank you so much for the changes! For the remaining requested changes (unresolved comments), did you address them as well or shall we discuss them more? I have...

Hi @chrisemezue, can you try a merge, now that Joey v2.0 is out?

Hi @chrisemezue can you try to resolve these? https://github.com/joeynmt/joeynmt/pull/164/conflicts And test if it runs?

Hi @NimaBoscarino thanks for chiming in. Yes, there are still conflicts. Not sure why the url doesn't work, it does on my end. I got there by clicking the button...

Thanks for adding the additional files @Ari-Ramkilowan! Do you have a link to the checkpoint of a trained model?

Hi @nikisix ! It looks like JW300 which we used as source for other languages does not include Pulaar. On the OPUS website you can look for other corpora: https://opus.nlpl.eu/...

That's super cool, thanks for sharing @kpu! What's the license for this data?

Looks good! Great work, @salomeyosei! :100: As far as I can tell, the only thing that needs to get fixed are the additional notebooks not under the benchmarks directory. Could...