Comment by mcemilg

I believe that if Mistral is serious about advancing in open source, they should consider sharing the corpus used for training their models, at least the base models pretraining data.