Yufan

Hi, batch size 1 might be too small, how many iterations did you use? The training code is for pre-training on a large dataset. If you want customize with a...

Hi, In promptnet.py, there are some layers that can output residual outputs which can be injected in to Stable Diffusion, similar to what ControlNet does. Those [residual outputs](https://github.com/drboog/ProFusion/blob/87d27607c7b47b6222560b73566ebeee92c4f891/train.py#L799) are currently...