differences from the first version of the model: this version takes up less space the training was carried out on a large number of photos the model turned out to be more responsive to prompts.