Questions on training Edge2Image and Sketch2Image

[alex4727]

Hi! I appreciate your great work.

I have question regarding training recipes of Sketch2Image and Edge2Image models.

1. What was the reason to train Sketch2Image from resumed checkpoint of Edge2Image? (I also checked that lora_rank for unet is increased from 8 to 128. How were there weights of lora rank 128 initialized?)
2. Do you plan to release the full training script for both cases?

Thanks!