Software Engineering Courses (SECourses)•13mo ago

dreambooth and fine tune are using same script

Furkan Gözükara SECoursesOP•12/23/24, 3:14 PM

only dataset preprationg slightly changes

Furkan Gözükara SECoursesOP•12/23/24, 3:14 PM

dreambooth without reg images = fine tuning

TTheToday Is there any news on Lora's training for Hunyuan?

Furkan Gözükara SECoursesOP•12/23/24, 3:14 PM

i didnt have chance to look at yet

Tonnu•12/24/24, 6:34 AM

Anyone have experience creating a finetuned model that is trained on a logo and can generate apparel / clothing with that logo?

Timson•12/24/24, 11:07 AM

L40 only 17% slower than L40S at fine tuning, really impressed

Timson•12/24/24, 11:26 AM

Looking at this: tips-info, I assumed it would be terrible

TTonnu Anyone have experience creating a finetuned model that is trained on a logo and ...

Furkan Gözükara SECoursesOP•12/24/24, 11:46 AM

usually not working that way great. but if you train logo you can apply it on random stuff

Furkan Gözükara SECoursesOP•12/24/24, 11:46 AM

or you can try in context lora of flux

TTimson L40 only 17% slower than L40S at fine tuning, really impressed

Furkan Gözükara SECoursesOP•12/24/24, 11:46 AM

amazing

TTimson Looking at this: https://discord.com/channels/772774097734074388/106779985541570...

Furkan Gözükara SECoursesOP•12/24/24, 11:48 AM

it could be either that guy didnt do accurate testing or it was runpod low quality pods :d

D4UNTLESS•12/24/24, 1:04 PM

For flux lora training on kohya, do I need to convert my webp images to png before starting the training? Or are webp images also supported?

DD4UNTLESS For flux lora training on kohya, do I need to convert my webp images to png befo...

Furkan Gözükara SECoursesOP•12/24/24, 1:11 PM

i didnt test but i expect to be supported. still best is png

FFurkan Gözükara SECourses i didnt test but i expect to be supported. still best is png

D4UNTLESS•12/24/24, 1:14 PM

Thanks I'll try with webp. Also, my flux training was successful. Only problem is that I cannot extract lora from it for some reason. I tried it on my local kohya installation which could be the problem. So I'll try to extract it now on cloud

DD4UNTLESS Thanks I'll try with webp. Also, my flux training was successful. Only problem i...

Furkan Gözükara SECoursesOP•12/24/24, 2:12 PM

could be

FFurkan Gözükara SECourses it could be either that guy didnt do accurate testing or it was runpod low quali...

Timson•12/24/24, 2:23 PM

nope, 31% slower at inference

strange

FFurkan Gözükara SECourses could be

D4UNTLESS•12/24/24, 2:45 PM

Do you have an updated method to extract lora from a trained flux model?

TTimson nope, 31% slower at inference 🤔 strange

Furkan Gözükara SECoursesOP•12/24/24, 5:24 PM

training is not that great optimized maybe that is why

DD4UNTLESS Do you have an updated method to extract lora from a trained flux model?

Furkan Gözükara SECoursesOP•12/24/24, 5:24 PM

no it is same as on patreon

Furkan Gözükara SECoursesOP•12/24/24, 5:24 PM

https://www.patreon.com/posts/how-to-extract-112335162

Patreon

How to Extract LoRA from FLUX Fine Tuning / DreamBooth Training Ful...

Get more from SECourses: FLUX, Tutorials, Guides, Resources, Training, Scripts on Patreon

FFurkan Gözükara SECourses dreambooth and fine tune are using same script

Yobin•12/25/24, 6:29 AM

ah noted on this! I've been doing fine tuning then lol which is better tho? DB or fine tune? I have 1000 images

YYobin ah noted on this! I've been doing fine tuning then lol which is better tho? DB o...

Furkan Gözükara SECoursesOP•12/25/24, 10:40 AM

DB = Fine tuning when no regularization images used. and for flux we dont use reg images since it doesnt work.

FFurkan Gözükara SECourses hi follow this video every step : https://youtu.be/0t5l6CP9eBg

ctz00•12/25/24, 3:15 PM

it worked thank you so much

D4UNTLESS•12/25/24, 4:59 PM

D4UNTLESS•12/25/24, 4:59 PM

I'm on SwarmUI

Cctz00 it worked thank you so much 😊

Furkan Gözükara SECoursesOP•12/25/24, 7:25 PM

awesome

DD4UNTLESS Click to see attachment

Furkan Gözükara SECoursesOP•12/25/24, 7:25 PM

try restart

Vigilence•12/25/24, 8:19 PM

Hi @Furkan Gözükara SECourses Is it still not possible to train 3.5?

VVigilence Hi <@205854764540362752> Is it still not possible to train 3.5?

Furkan Gözükara SECoursesOP•12/25/24, 8:19 PM

i think should be but i didnt have chance yet

Vigilence•12/25/24, 8:19 PM

@Furkan Gözükara SECourses What would we use?

VVigilence <@205854764540362752> What would we use?

Furkan Gözükara SECoursesOP•12/25/24, 8:25 PM

kohya supports as far as i know

Furkan Gözükara SECoursesOP•12/25/24, 8:25 PM

or check out onetrainer

Vigilence•12/25/24, 8:29 PM

@Furkan Gözükara SECourses Do you still have plans to do full finetune plans?

VVigilence <@205854764540362752> Do you still have plans to do full finetune plans?

Furkan Gözükara SECoursesOP•12/25/24, 8:33 PM

yes if i can find time hopefully

Vigilence•12/25/24, 8:34 PM

@Furkan Gözükara SECourses I look forward to it. I haven't been here for a month so I thought some progress was made.

VVigilence <@205854764540362752> I look forward to it. I haven't been here for a month so ...

Furkan Gözükara SECoursesOP•12/25/24, 9:26 PM

i agree

Timson•12/26/24, 3:38 PM

@Furkan Gözükara SECourses have you tried to extract loras from slightly overtrained checkpoints instead of increasing lora weight? Does it yield results closer to fine tuning with lower (best) training result?

TTimson <@205854764540362752> have you tried to extract loras from slightly overtrained ...

Furkan Gözükara SECoursesOP•12/26/24, 5:45 PM

i didnt but sounding good in theory

Renzo•12/27/24, 4:16 PM

@Furkan Gözükara SECourses, I'm failing to remember what is the formula you use to calculate the learning rate based on the batch size. Could you share it?

RRenzo <@205854764540362752>, I'm failing to remember what is the formula you use to ca...

Furkan Gözükara SECoursesOP•12/27/24, 5:11 PM

square root of batch size * leraning rate

Furkan Gözükara SECoursesOP•12/27/24, 5:12 PM

NiuNiuKing_AI•12/28/24, 3:20 PM

@Furkan Gözükara SECourses , is there a tutorial or configuration on how to do Textual inversion in kohya_ss?

NNiuNiuKing_AI <@205854764540362752> , is there a tutorial or configuration on how to do Textua...

Furkan Gözükara SECoursesOP•12/28/24, 3:31 PM

i dont have any atm. textual inversion is least powerful one

NiuNiuKing_AI•12/28/24, 3:35 PM

thx, yes but because it is light, it is preferred to combine with lora in many settings. Since you have so much experience in kohya_ss, is this something you can test easily?

NNiuNiuKing_AI thx, yes but because it is light, it is preferred to combine with lora in many s...

Furkan Gözükara SECoursesOP•12/28/24, 4:21 PM

well i had tested it way back

Furkan Gözükara SECoursesOP•12/28/24, 4:21 PM

https://www.youtube.com/watch?v=dNOpWt-epdQ&t=4s

YouTubeSECourses

How To Do Stable Diffusion Textual Inversion (TI) / Text Embeddings...

Our Discord : https://discord.gg/HbqgGaZVmr. Grand Master tutorial for Textual Inversion / Text Embeddings. 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

https://www.patreon.com/SECourses

Playlist of Stable Diffusion Tutorials, Automatic1111 and Google Colab Guide...

How To Do Stable Diffusion Textual Inversion (TI) / Text Embeddings...

Furkan Gözükara SECoursesOP•12/28/24, 4:21 PM

but not with kohya

dawnslayerz•12/28/24, 10:20 PM

@Furkan Gözükara SECourses i followed the steps in your video for runpod install of kohya for flux lora training and when i open kohya there is no "flux" portion under peremeters. not sure why. i copied your video exactly.

Ddawnslayerz <@205854764540362752> i followed the steps in your video for runpod install of k...

Furkan Gözükara SECoursesOP•12/28/24, 10:32 PM

it will appear when you give accurate folder path of the flux dev model

dawnslayerz•12/28/24, 10:36 PM

@Furkan Gözükara SECourses oh, ok, ty

ahmedjamal9673•12/29/24, 6:35 AM

@Furkan Gözükara SECourses
Hi Doctor
Could you kindly please send me the settings text file of flux tools on swarmui as i work on kaggle?

dreambooth and fine tune are using same script

Similar Threads

Similar Threads

Similar Threads