Make the preference data for text-to-image models available?

#7
by sayakpaul HF staff - opened

Would be very nice to have the preference dataset available. It will help foster the research in aligning diffusion models further. In my understanding, the only viable option right now for this is: https://huggingface.co/datasets/yuvalkirstain/pickapic_v2.

Create a full round logo with lighthouse, fish and sailing boat elements in different colours on a black background. The logo should read "GELIBOLU BELEDIYESI 1923" in a prominent and large writing style in 3D view.

2LlWjL-ISM29oHxtbdrvXw.png

TIGER-Lab org

We will release it once we hit 5000 votes

TIGER-Lab org
wenhu changed discussion status to closed
This comment has been hidden

Sign up or log in to comment