* Add a vision transformer example (dino-v2). * Add some documentation + test. * CI fix. * Another fix (still unable to replicate the errors locally :( )