HuggingFace integration reference image mode works significantly worse than original repository
#6
by
maxtes
- opened
Hi! Your integration works fine in text prompt mode, but when it comes to reference image mode, model's predictions look inadequate in comparison to predictions of original model on the same image with the same settings. You can check predictions of original model here.
I will attach an image for comparison to this comment.
yeah I noticed the same.
Any updates on this?
