ValueError: `resolution_max_side` cannot be larger than `max_image_size` with N=5

#9
by rtbonet - opened

I tried to set N to 5 or even 6 as described for a 3000x3000 image, like below:
inputs = processor(text=prompt, images=[image], size={"longest_edge": 5*384}, return_tensors="pt")
But this error pops up:
ValueError: resolution_max_side cannot be larger than max_image_size

Without setting longest_edge, the output just repeats the same word when dealing with an image that represents a big table.

Hugging Face TB Research org

Hi! I actually fixed this last week in transformers. It's there on main but they haven't made a new release since my fix

Sign up or log in to comment