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.