Comment by tippytippytango

The python zoom in seems performative. A vision model already has access to all the data, how does zooming in help it? Still very cool that it can!