Datasets:
More information about translation process
#2
by stefan-it - opened
Hi @ezosa ,
this translated dataset is a really cool new resource. I would like to know a few more details about the translation process. So e.g. which Dolci dataset is used, is it this one?
And can you give more details about the translation process, e.g. which model is used + translation prompt and the max. context size. Does this dataset also include multi-turn conversations and thinking traces - were thinking/reasoning traces also translated 🤔
It seems that for every id, translated intructions do exists for each language - are there exceptions? And my last question would be: is there a reason why "only" a subset from the original Dolci dataset is used, are any pre-filtering criteria used?
Many thanks in advance!