Software Engineering Courses (SECourses)•3y ago

I want to create a style LORA based on a 1.5 model, I tried to follow your advice here to get Koyha

I want to create a style LORA based on a 1.5 model, I tried to follow your advice here to get Koyha working, but I am using Runpod, do you know which file I need to edit to get a Gradiolink to come up?

JJas I want to create a style LORA based on a 1.5 model, I tried to follow your advic...

Furkan Gözükara SECourses•10/12/23, 5:36 PM

please use my kohya installer on runpod it has instrucitons too : https://www.patreon.com/posts/for-runpod-kohya-84898806

FFurkan Gözükara SECourses check this out guys : https://twitter.com/GozukaraFurkan/status/1712518063256449...

Elisoli•10/12/23, 5:50 PM

for me blip2-opt-6.7b-coco was the worst, very different prompts from analyzing my photos

FFurkan Gözükara SECourses please use my kohya installer on runpod it has instrucitons too : https://www.pa...

JasOP•10/12/23, 5:52 PM

is this specific for SDXL lora? I would like to train a 1.5 lora

FFurkan Gözükara SECourses check this out guys : https://twitter.com/GozukaraFurkan/status/1712518063256449...

Elisoli•10/12/23, 6:00 PM

ModuleNotFoundError: No module named 'clip_interrogator'
where to i get this?

FFurkan Gözükara SECourses or transparent background

Benjamin•10/12/23, 6:12 PM

removing the background improves the training?

SStefanoid9343 Hi @Dr. Furkan Gözükara I have the following message on kaggle and I cannot type...

t800se•10/12/23, 6:16 PM

I managed to fix it by replacing the !apt update with !apt-get update --allow-releaseinfo-change in the beginning of that cell

JJas is this specific for SDXL lora? I would like to train a 1.5 lora

Furkan Gözükara SECourses•10/12/23, 6:51 PM

it is for installing kohya

Furkan Gözükara SECourses•10/12/23, 6:51 PM

you failed to launch it

EElisoli ModuleNotFoundError: No module named 'clip_interrogator' where to i get this?

Furkan Gözükara SECourses•10/12/23, 6:52 PM

did you run install file?

BBenjamin removing the background improves the training?

Furkan Gözükara SECourses•10/12/23, 6:52 PM

better than white i think

Tt800se I managed to fix it by replacing the **!apt update** with **!apt-get update --al...

Furkan Gözükara SECourses•10/12/23, 6:52 PM

i also updated the file

FFurkan Gözükara SECourses better than white i think

Benjamin•10/12/23, 6:52 PM

ah ok, thought in general

t800se•10/12/23, 7:02 PM

I'm getting this error now

FFurkan Gözükara SECourses did you run install file?

Elisoli•10/12/23, 7:04 PM

yes but before this new update to the patreon

Elisoli•10/12/23, 7:10 PM

virginia•10/12/23, 7:10 PM

@Dr. Furkan Gözükara also got it, let me try again

FFurkan Gözükara SECourses please use my kohya installer on runpod it has instrucitons too : https://www.pa...

JasOP•10/12/23, 7:14 PM

I am using the Ashleykza template which has Koyha installed, but I was worried that the trainings are not currently good, due to the update you posted on Patreon.

Or am I getting confused and this does not affect trainings?

Tt800se I'm getting this error now 😦

Furkan Gözükara SECourses•10/12/23, 7:16 PM

use v4 file

Furkan Gözükara SECourses•10/12/23, 7:16 PM

redownload please

EElisoli Click to see attachment

Furkan Gözükara SECourses•10/12/23, 7:16 PM

please try reinstall

Furkan Gözükara SECourses•10/12/23, 7:16 PM

i am testing too

FFurkan Gözükara SECourses use v4 file

t800se•10/12/23, 7:17 PM

will do now, thank you

EElisoli yes but before this new update to the patreon

Furkan Gözükara SECourses•10/12/23, 7:17 PM

you are using all_files.zip right?

Furkan Gözükara SECourses•10/12/23, 7:18 PM

preparing runpod version of captioner

JJas I am using the Ashleykza template which has Koyha installed, but I was worried t...

Furkan Gözükara SECourses•10/12/23, 7:19 PM

i am not sure. i havent tested. but i am still searching for a better workflow for sd 1.5

FFurkan Gözükara SECourses i am not sure. i havent tested. but i am still searching for a better workflow f...

JasOP•10/12/23, 7:23 PM

ok, I figured out what to do, just had to change PUBLIC=false to PUBLIC=true in the setup.sh file

t800se•10/12/23, 7:43 PM

Ok, I am not getting that error with the v4, thank you @Dr. Furkan Gözükara . I see there is a new cell in this version, when shall I execute that one?

Tt800se Ok, I am not getting that error with the v4, thank you @Dr. Furkan Gözükara . I ...

Furkan Gözükara SECourses•10/12/23, 7:44 PM

it is optional

t800se•10/12/23, 7:44 PM

ahh good to know

Furkan Gözükara SECourses•10/12/23, 7:54 PM

runpod installer added

Furkan Gözükara SECourses•10/12/23, 7:54 PM

https://www.patreon.com/posts/90744385

Furkan Gözükara SECourses•10/12/23, 7:54 PM

still testing

SuperTurboHero•10/12/23, 9:20 PM

i must be doing something wrong

it's RealisticVision5.1, Dreambooth trained. on do you guys know any good tutorials where I can learn how to get good looking realistic photos?

SuperTurboHero•10/12/23, 9:25 PM

the best results so far I got from Automatic1111 v1.4.0 interface, because there are these options "restore faces" and "Hires.fix"

t800se•10/12/23, 9:26 PM

Hey @Dr. Furkan Gözükara , very silly question probably but I am new in all of these

I just started the training process in Kaggle Notebook by watching your videos. I didn't relise that it will take 7-9 hrs

Do I need to leave wifi on and my pc on while it trains? or because it's in the cloud, so it doesn't matter?

SSuperTurboHero i must be doing something wrong 😄 it's RealisticVision5.1, Dreambooth trained. ...

Furkan Gözükara SECourses•10/12/23, 11:21 PM

looks like training errors

Furkan Gözükara SECourses•10/12/23, 11:21 PM

what is your gpu 12gb?

Tt800se Hey @Dr. Furkan Gözükara , very silly question probably but I am new in all of t...

Furkan Gözükara SECourses•10/12/23, 11:22 PM

yes you need to leave browser open on kaggle

Furkan Gözükara SECourses•10/12/23, 11:22 PM

on runpod you dont have to

JasOP•10/12/23, 11:40 PM

If i train a dreambooth SDXL model on a style, is it weird to use regulisation images?

JasOP•10/12/23, 11:42 PM

i am making a model of images like this

JasOP•10/12/23, 11:43 PM

so it would be strange to have regularisation images of portrait photos of people?

JJas If i train a dreambooth SDXL model on a style, is it weird to use regulisation i...

Furkan Gözükara SECourses•10/12/23, 11:51 PM

you dont have to use

Furkan Gözükara SECourses•10/12/23, 11:51 PM

if you dont have good ones

Furkan Gözükara SECourses•10/12/23, 11:51 PM

i would do 2 trainings and compare

FFurkan Gözükara SECourses you dont have to use

JasOP•10/12/23, 11:59 PM

thanks for quick reply, from what I understand the regularisation images should be images that you would expect to see when using the class prompt?

So if i use the word "style" as a class prompt then should I have regularisation images of different styles?

Really the images will all be kind of macro photos of colourful water and such like this...

What kind of regularisation images would be best?

JJas thanks for quick reply, from what I understand the regularisation images should ...

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

the logic is like this

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

we are leveraging a class info knowladge

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

reg images should be as much as possibly like those class images used in the original training of model

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

so it depends on your class word

I want to create a style LORA based on a 1.5 model, I tried to follow your advice here to get Koyha

Similar Threads

Similar Threads

Similar Threads