Software Engineering Courses (SECourses)•3y ago

what should I do then? there is no way to use Kohya with a downgraded version, right?

LLe what should I do then? there is no way to use Kohya with a downgraded version, r...

Furkan Gözükara SECourses•6/6/23, 5:08 PM

there is a way certainly

Furkan Gözükara SECourses•6/6/23, 5:08 PM

https://github.com/FurkanGozukara/Stable-Diffusion/blob/main/Tutorials/Generate-Studio-Quality-Realistic-Photos-By-Kohya-LoRA-Stable-Diffusion-Training-Full-Tutorial.md

GitHub

Stable-Diffusion/Generate-Studio-Quality-Realistic-Photos-By-Kohya-...

Best Stable Diffusion and AI Tutorials, Guides, News, Tips and Tricks - Stable-Diffusion/Generate-Studio-Quality-Realistic-Photos-By-Kohya-LoRA-Stable-Diffusion-Training-Full-Tutorial.md at main · ...

Furkan Gözükara SECourses•6/6/23, 5:08 PM

do a new git clone

Furkan Gözükara SECourses•6/6/23, 5:08 PM

then do

Furkan Gözükara SECourses•6/6/23, 5:08 PM

git checkout 63657088f4c35a376dd8a936f53e9b9a3b4b1168

Furkan Gözükara SECourses•6/6/23, 5:08 PM

this is the version i used

Furkan Gözükara SECourses•6/6/23, 5:08 PM

that is the version i used in the video

LeOP•6/6/23, 5:09 PM

Ok, I will try.

LeOP•6/6/23, 5:09 PM

Thank you

Furkan Gözükara SECourses•6/6/23, 5:09 PM

which gpu do you have?

Ffelixsanz but inpaint using "only masked"

LeOP•6/6/23, 5:09 PM

I tried that too. I get a funko pop on inpaint

LeOP•6/6/23, 5:09 PM

:crying:

LeOP•6/6/23, 5:11 PM

The quality is impressive. I got amazing results for my face. But the LoRA is unresponsive. It doesn't respect prompt. If it does I get a saturated low-quality version of myself and ignores full-body requests. It is so annoying.

LeOP•6/6/23, 5:11 PM

Going to try that tutorial.

LLe The quality is impressive. I got amazing results for my face. But the LoRA is un...

felixsanz•6/6/23, 5:12 PM

probably overtrained

Ffelixsanz probably overtrained

LeOP•6/6/23, 5:12 PM

I used something likee 2000 steps

felixsanz•6/6/23, 5:13 PM

holy s, maybe stable diffusion goes crazy with so much beauty xd

felixsanz•6/6/23, 5:13 PM

i don't like loras, they are not flexible at all in my experience

LeOP•6/6/23, 5:13 PM

do you prefer TI?

felixsanz•6/6/23, 5:13 PM

no, that's like plastic results

felixsanz•6/6/23, 5:13 PM

dreambooth

Furkan Gözükara SECourses•6/6/23, 5:14 PM

ye dreambooth best

Furkan Gözükara SECourses•6/6/23, 5:14 PM

if you have a card

Furkan Gözükara SECourses•6/6/23, 5:14 PM

or hiring runpod

LeOP•6/6/23, 5:14 PM

I have a 3060

LeOP•6/6/23, 5:14 PM

12 GB

Furkan Gözükara SECourses•6/6/23, 5:14 PM

ye that is limiting

LeOP•6/6/23, 5:14 PM

felixsanz•6/6/23, 5:14 PM

runpod is cheap

Furkan Gözükara SECourses•6/6/23, 5:14 PM

you need to make a good lora

felixsanz•6/6/23, 5:14 PM

my 2070 is even more limiting XD

felixsanz•6/6/23, 5:15 PM

but with 25$ on runpod you can test and create a lot of models, like 30-50 experiments

LeOP•6/6/23, 5:15 PM

I dont mind waiting. I can wait for hours on dreambooth if it gives me nice results

LeOP•6/6/23, 5:15 PM

Oh, I am down to pay

felixsanz•6/6/23, 5:15 PM

maybe you can dreambooth on 12GB using xformers, but that reduces a bit of quality

LeOP•6/6/23, 5:16 PM

if it gives me nice results, I can do upscaler on img2img no problem

felixsanz•6/6/23, 5:16 PM

what if you try the dreambooth colab? it's free

LeOP•6/6/23, 5:16 PM

I need to google that. Never tried using cloud services, only on my pc.

felixsanz•6/6/23, 5:18 PM

https://colab.research.google.com/github/FurkanGozukara/Stable-Diffusion/blob/main/DreamBooth/ShivamShriraoDreamBooth.ipynb

felixsanz•6/6/23, 5:19 PM

there is a video for this, i think

LeOP•6/6/23, 5:19 PM

https://www.youtube.com/watch?v=zA4LksIVas8

YouTubeSECourses

RunPod Fix For DreamBooth & xFormers - How To Use Automatic1111 Web...

RunPod: https://bit.ly/RunPodIO. Discord : https://bit.ly/SECoursesDiscord. How to install latest version of Automatic1111 Web UI and #DreamBooth on #RunPod. If I have been of assistance to you and you would like to show your support for my work, please consider becoming a patron on