More fix for tiny model creation (#45228)
* add missing VivitImageProcessor
* fix model_type = "evolla"
* change to google/medasr
* fix auto_docstring tue-mps/videomt-dinov2-small-ytvis2019
* fix 3 tester
* fix test
* fix modular
* fix modular
---------
Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>