Software Engineering Courses (SECourses)•3y ago

Does anyone have thoughts on how I can train my personal LoRA to not include the background? I know

Does anyone have thoughts on how I can train my personal LoRA to not include the background? I know part of this is using an earlier epoch, however before/during the training process is there anything I can do to really ensure it only trains the person?

Sshortdwarf Does anyone have thoughts on how I can train my personal LoRA to not include the...

strakovic•12/12/23, 8:34 PM

Remove background

shortdwarfOP•12/12/23, 8:35 PM

I considered swapping with plain color backgrounds (ex. one red, one blue, etc) or would white be best?

Sshortdwarf I considered swapping with plain color backgrounds (ex. one red, one blue, etc) ...

strakovic•12/12/23, 8:36 PM

Personally don’t see a benefit using other than white

shortdwarfOP•12/12/23, 8:37 PM

great, i only hestitated doing this because I thought it might lead to a strong constrast between subject and background but ill give it a shot. is there a program youd recommend using to swap in the white background or just photoshop?

Sshortdwarf Does anyone have thoughts on how I can train my personal LoRA to not include the...

Furkan Gözükara SECourses•12/12/23, 8:37 PM

using different backgrounds

Sshortdwarf I considered swapping with plain color backgrounds (ex. one red, one blue, etc) ...

Furkan Gözükara SECourses•12/12/23, 8:37 PM

it learns this

FFurkan Gözükara SECourses it learns this

shortdwarfOP•12/12/23, 8:38 PM

so do you think if i used simple backgrounds of plain colors it wouldnt learn it as much if there was variety of colors?

Sshortdwarf so do you think if i used simple backgrounds of plain colors it wouldnt learn it...

Furkan Gözükara SECourses•12/12/23, 8:57 PM

ye plain backgrounds were not better

Furkan Gözükara SECourses•12/12/23, 8:57 PM

in my sd 1.5 test

Furkan Gözükara SECourses•12/12/23, 8:57 PM

but dont know for sdxl

shortdwarfOP•12/12/23, 8:59 PM

okay, ill run some tests on plain white, plain multi colors, and anything else that comes to mind and share what i find

Kallamamran•12/12/23, 9:07 PM

Hey @Dr. Furkan Gözükara I have Python 3.10.9.... Should I somehow update to 3.10.11 or am I fine with 3.10.9?

KKallamamran Hey @Dr. Furkan Gözükara I have Python 3.10.9.... Should I somehow update to 3....

Furkan Gözükara SECourses•12/12/23, 9:07 PM

3.10.9 is fine too

Kallamamran•12/12/23, 9:12 PM

Good.

Kallamamran•12/12/23, 9:12 PM

How do I change the subject from person to car in SOTA?

Kallamamran•12/12/23, 9:13 PM

There?

KKallamamran There?

Furkan Gözükara SECourses•12/12/23, 9:31 PM

yes

FFurkan Gözükara SECourses yes

Kallamamran•12/12/23, 9:46 PM

Great

Kallamamran•12/12/23, 9:56 PM

If the aspect ratios is just "aspect ratios", why are they specified as for example [1536, 1536] instead of [1, 1]?

KKallamamran If the aspect ratios is just "aspect ratios", why are they specified as for exam...

Furkan Gözükara SECourses•12/12/23, 10:07 PM

well to understand easier

Furkan Gözükara SECourses•12/12/23, 10:08 PM

but same as 1:1

Furkan Gözükara SECourses•12/12/23, 10:08 PM

check out my LLaVA captioning

Furkan Gözükara SECourses•12/12/23, 10:08 PM

https://www.linkedin.com/posts/furkangozukara_llava-16-bit-batch-image-captioning-is-using-activity-7140462239316180992-HgHd?utm_source=share&utm_medium=member_desktop

Furkan Gözükara on LinkedIn: LLaVA 16 bit batch image captioning is...

LLaVA 16 bit batch image captioning is using 27729 MB VRAM

Installed via my Auto RunPod installer

I am captioning style images for a new project for a…

LLorelai AI I spend more time fixing and trying to install broken TensorRT than the time I'd...

mtnmecca•12/12/23, 10:26 PM

my experience as well. on first start up of a1111 I can get tensorRT to work. after that it just gets stuck on errors. It does not seem reliable yet. I also spend too much time trying to get past the errors.

Mmtnmecca my experience as well. on first start up of a1111 I can get tensorRT to work. ...

Furkan Gözükara SECourses•12/12/23, 10:36 PM

we solved his problem working now

୧

୧ʕ•̀ᴥ•́ʔ୨•12/12/23, 10:42 PM

Can anyone suggest a good anti-nsfw/adult negative prompt?

୧

୧ʕ•̀ᴥ•́ʔ୨•12/12/23, 10:44 PM

Not sure what to do to ensure adult stuff doesn't pop up when generating people, don't want to get banned again

୧୧ʕ•̀ᴥ•́ʔ୨ Not sure what to do to ensure adult stuff doesn't pop up when generating people,...

Furkan Gözükara SECourses•12/12/23, 10:54 PM

i think depends on your base model

Furkan Gözükara SECourses•12/12/23, 10:55 PM

some models are too nasty

Furkan Gözükara SECourses•12/12/23, 10:55 PM

maybe extension can block

Furkan Gözükara SECourses•12/12/23, 10:55 PM

generated nsfw image

Furkan Gözükara SECourses•12/12/23, 10:58 PM

try https://github.com/AUTOMATIC1111/stable-diffusion-webui-nsfw-censor

GitHub

GitHub - AUTOMATIC1111/stable-diffusion-webui-nsfw-censor: stable-d...

stable-diffusion-webui-nsfw-censor. Contribute to AUTOMATIC1111/stable-diffusion-webui-nsfw-censor development by creating an account on GitHub.

Mmtnmecca my experience as well. on first start up of a1111 I can get tensorRT to work. ...

Lorelai AI•12/12/23, 11:00 PM

Yes it's working great. I'm doing some other things but here's some items to consider. Loras that aren't integarted into the TensorRT engine don't work. You can add them but they dont work usefully even at high Lora weights. My project tonight as @Dr. Furkan Gözükara sugested is im going to merge 1-2 Loras into a checkpoint and then make an engine

୧୧ʕ•̀ᴥ•́ʔ୨ Can anyone suggest a good anti-nsfw/adult negative prompt?

Lorelai AI•12/12/23, 11:10 PM

I need to make a facial expression slider lora, that is a big to do on my list

LLorelai AI I need to make a facial expression slider lora, that is a big to do on my list

୧

୧ʕ•̀ᴥ•́ʔ୨•12/12/23, 11:32 PM

Or a dropdown where you choose an expression type? Or emotion?

୧

୧ʕ•̀ᴥ•́ʔ୨•12/12/23, 11:37 PM

A slider would work for something like a smile/frown amount, as it's mostly 1-dimensional

୧୧ʕ•̀ᴥ•́ʔ୨ Or a dropdown where you choose an expression type? Or emotion?

Lorelai AI•12/12/23, 11:39 PM

I actually am thinking of just making it a general lora. I'll caption my dataset with the expressions and have a good amount of subjects

୧

୧ʕ•̀ᴥ•́ʔ୨•12/12/23, 11:42 PM

I'm surprised this doesn't exist yet (or maybe it does)

୧

୧ʕ•̀ᴥ•́ʔ୨•12/12/23, 11:49 PM

Maybe people tried but it's not trivial and doesn't work very well

Lorelai AI•12/12/23, 11:56 PM

It does but I haven’t found one in Sdxl with the multiple expressions. It’s per expression. I also hypothesize if I get better in my captions and label expressions my expression prompts will be better for those models

Lorelai AI•12/13/23, 12:17 AM

I'm going to make one right now I have an idea

Lorelai AI•12/13/23, 12:22 AM

actually haha, lot more work than I expected

Bophedes•12/13/23, 1:05 AM

Hey I got an issue downloading stable-diffusion using the automatic installer. is this where i ask for help?

Furkan Gözükara SECourses•12/13/23, 1:10 AM

https://www.linkedin.com/posts/furkangozukara_started-3-style-trainings-for-a-client-activity-7140507991744720897-ShL2?utm_source=share&utm_medium=member_desktop

Furkan Gözükara on LinkedIn: Started 3 style trainings for a client...

Started 3 style trainings for a client.

Using RunPod RTX 4090.

Training on SDXL 1.0 base version with Text Encoder as well.

Using my very best…

BBophedes Hey I got an issue downloading stable-diffusion using the automatic installer. i...

Furkan Gözükara SECourses•12/13/23, 1:10 AM

hello

Furkan Gözükara SECourses•12/13/23, 1:11 AM

which link are you using?

Bophedes•12/13/23, 1:11 AM

https://github.com/EmpireMediaScience/A1111-Web-UI-Installer/releases/tag/V1.7.0

i am using this one

GitHub

Release A1111 Webui Launcher V1.7.0 · EmpireMediaScience/A1111-Web-...

Adding an "no-autolaunch" launch option for the launcher, just add it at the end of the target of the "A1111 WebUI (Pin to Taskbar)" shortcut, like the "skip" launch o...

BBophedes https://github.com/EmpireMediaScience/A1111-Web-UI-Installer/releases/tag/V1.7.0...

Furkan Gözükara SECourses•12/13/23, 1:11 AM

use mine

Furkan Gözükara SECourses•12/13/23, 1:11 AM

https://www.patreon.com/posts/86307255

Furkan Gözükara SECourses•12/13/23, 1:11 AM

if you are not my patreon supporter : https://youtu.be/kvxX6NrPtEk?si=-mXu6rXWsG1IgczZ

YouTubeSECourses

Double Your Stable Diffusion Inference Speed with RTX Acceleration ...

Stable Diffusion Gets A Major Boost With RTX Acceleration. One of the most common ways to use Stable Diffusion, the popular Generative AI tool that allows users to produce images from simple text descriptions, is through the Stable Diffusion Web UI by Automatic1111. In today’s Game Ready Driver, NVIDIA added TensorRT acceleration for Stable Diff...

Does anyone have thoughts on how I can train my personal LoRA to not include the background? I know

Similar Threads