Can you share the size of the datasets, and some parameter used for training? I'm trying to understand what is the minimum size of a corpus and the architectur

Can you share the size of the datasets, and some parameter used for training? I'm trying to understand what is the minimum size of a corpus and the architecture size too.