Training SDXL lora with OneTrainer

First: very, very thanks to

Stable Cascade Lora Training with OneTrainer | Civitai

At first run system have to dowload a lot of data...

For me, it not working and first time... so i try again and again... and finally:

So, lets go!

Put json (presets file) setting to folder

Step by step:

I have images and txt files on drive L, but you can set everywhere.

I use stabilityai format od diffusion model base:

In Onetrainer you dont have REPEATS option, so you need to make copy of your concept and make it active. Just press "+"

I was use dataset tool to make captions with BLIP and set trigger word as first in caption.

If you need longer caption you have to add max_tokens in blip or WD14.

and last one window

In progress.......

You can sample in any time (turn off automaticly)

Testing workflow:

And normal workflow

no lora

lora 0.1

lora 0.9

Have a nice day!