Training details

#3
by Qin-YL - opened

Hello, what are the specific training details?

Owner

I used around 30k images generated by the parent model along with midas depth maps of all of the images. I then made a prediction with the ipadapter bypassed but using a depth t2i adapter. I use this prediction for my target. Let me know if that is not clear and I can expand on it a bit more.

hi, would you give a little bit more details...

Sign up or log in to comment