how many classes can I input in one text?

#5
by xingwang1234 - opened

I noticed "the maximum text token number as 256", so it means the input class text length (e.g. cat. remote control. xxx. xxx. xxx.....) is very limited, right?

@xingwang1234 Yeah we're limited to 256 text tokens in case you have more than that in your prompt you'd have to break your prompt and run inference more than one time. This is a limitation of GroundingDino other models like. It was my plan to add to the transformers library a model called APE that deals with this problem, but unfortunately I don't have the time to do it anymore and the reviews were taking way to long as well :(

@EduardoPacheco Alright, understood. Thanks for the amazing work on GroundingDino.

Sign up or log in to comment