VILA icon indicating copy to clipboard operation
VILA copied to clipboard

Does S2 able to unfreeze vit to train?

Open MonolithFoundation opened this issue 1 year ago • 1 comments

I think if using s2, and unfreeze vit, the result could be worse, as the s2 split images.

MonolithFoundation avatar May 17 '24 03:05 MonolithFoundation

Hi, the results of VILA-3B-S2 is trained with ViT unfrozen. We didn't observe any negative effect of that.

bfshi avatar May 20 '24 05:05 bfshi