Concern about the performance while using multi image vs single image inference
#66
by
wiccanmind
- opened
As my experience with this 7B version, given
[Image_A] which contain the [image_a1] and [image_a2], and a set of question about each [image_a1] and [image_a2], respectively.
I realize that the model work better if I feed the big image [Image_A], instead of using multi image (image_a1 and image_a2)
If this observation is right, can anyone explain the technique issue behind? Thanks <3