Efficient-Large-Model

/

VILA-7b

like 20

Is there a limit to how many images can be supplied as context

First, please use VILA1.5 models. that will give you better performance
Second, for VILA1.5 models, in theory you can fit in 20 images for 8B (196 tokens/image and 4096 context length) but we only tested in-context learning w/ 2 images and video input w/ 8 frames.

Thank you for the information 😊