Dolly Machine Learning

We have several hundred million paired images and captions from the interne. A large portion of this isjust data that we've purchased then and licensed. Of course, there's some eden from other languages. But ntr, for this eneration, we try to focus on english.