search

ZichengDuan / TheChosenOne

Unofficial implementation of the paper "The Chosen One: Consistent Characters in Text-to-Image Diffusion Models"
https://arxiv.org/abs/2311.10093
242 stars 22 forks source link

Questions about the training code part #14

Closed just1nseo closed 9 months ago

just1nseo commented 9 months ago

Hi. Thank you for your implementation of the paper.

While I was looking at your code, I couldn't understand whether the model being trained each loop is the model that was trained in the previous loop, or you are calling a vanilla SDXL model every loop.

Can you tell me what is right, and where the appropriate code for my question is ?

Marstaos commented 9 months ago

I think it's the model from the previous round of the loop. In the paper, it is said that continuous training in multiple rounds finally achieves the desired effect