IPAdapter Support #283
Description
Hi, I have tried IPAdapter experiment depends on SANA, but the result is terrible, the training cant be converged and the loss and grad_norm gone nan after about 4000 steps. Moreover, I have also tried Uno (https://bytedance.github.io/UNO) experiment, the result is also terrible, maybe there is something wrong with my code. Do you have some plan to publish the IPAdapter version? or give some advice?
Thank you for your great work! I think the project will be more popular if some assistant work(like Img2Img, IPAdapter, Inpaint and so on) be support.