Stable Diffusion 0.58 with Dreambooth training (Patreon)

Hi all.
Ok, first of all, dreambooth training use a lot of crazy new libraries to train it under 12vram. So if the code runs in this first version, we already are very lucky lol

In my last post I did say that the GUI would be separated download from the models. I was not able to make this change in this version, since Dreambooth was a lot of work, but I'm already working to make this change on the next build.

There is also a new option.
Clip Skip: This pretty much use other layers of the Text encoder model. It make the results care less for details and more for the overall feeling on the image.

NovelAI use Clip Skip=1 for training, I think it work a little better for anime. It seen that the bigger value it give more dreamy results.
The default value is 0 (same as the older version)

Training with Dreambooth:
Here is the tutorial that I will keep updating later on:
https://docs.google.com/document/d/15gsTo2IaNxSfGox_RQatg1nuIyxs3Tm0fyUZMUHTXbw/edit?usp=sharing 

Link to the build:

https://drive.google.com/file/d/1iXsPLbqeS-G8_1FxsLuevfaAdViDvxrO/view?usp=share_link 


Light Build:

Downloading this version, you need to copy the models and torch folder from the older build, since you don't need to download them again. Also Upscale will not work on this version. (Update work on the other build)

This is not trully the Light build version yet, that is still being worked on, this is just a temp solution for now.
https://drive.google.com/file/d/1Jg4hk1PVW_Cxkzxfq0k8iiAaFnLb2TMM/view?usp=share_link