Tips & Tricks ‐ LoRA Settings - InfluxOW/Stable-Diffusion-Text-To-Person GitHub Wiki

If you've found a great LoRA and you want to know settings it has been trained on, there's no need to ask the author! Sometimes it's possible to find its training settings in the model metadata. You just need to import it in the AUTOMATIC 1111. Then open Lora tab. Above some models you'll see info button. If it's there, then you can see the model internal metadata.

You can't just import it into Kohya SS or anywhere else. But it's pretty easy to match these settings with Kohya SS settings. As you could see, there's a lot of information: checkpoint, resolution, clip skip, number of training and regularisation images, number of steps, LoRA type, scheduler, learning rates and everything else. In general, there's everything you need! Just fill Kohya SS settings one by one, and you'll be fine.