The number of tokens *per image* are actually fairly small, ranging from 85 to ~...

		diwank on July 10, 2024 \| parent \| context \| favorite \| on: Vision language models are blind The number of tokens per image are actually fairly small, ranging from 85 to ~500.