Software Engineering Courses (SECourses)•14mo ago

I am registered and I have the Json's included in the LoRA_Tab_LoRA_Training_Best_FLUX_Configs folde

I am registered and I have the Json's included in the LoRA_Tab_LoRA_Training_Best_FLUX_Configs folder, but I think they are more oriented to train human figure or characters. Which one could I use for my purpose of creating Lora for still life?

dxqb•10/31/24, 10:32 PM

did you make this @Furkan Gözükara SECourses https://discord.com/channels/696818921625485342/1252662832076685334/1300976702880481291 ?

LLeolis78 I am registered and I have the Json's included in the LoRA_Tab_LoRA_Training_Bes...

wardensc2•11/1/24, 12:57 AM

Read more carefully, he already show how to train person and train style. There is no shortcut to it. There are no perfect settings. It depends on your dataset as well and also experiments. He's already shown how to do experiment how to check quality. Perfect settings is judged by you not him he only shows the way to do it. Get to work you lazy people

masood•11/1/24, 7:38 AM

hey everyone, i want to do Dreambooth training of flux-dev-de-destill. I am a beginner and will really appreciatesuggestions. Thanks

dxqb•11/1/24, 8:30 AM

@Furkan Gözükara SECourses not sure if this is interesting for you, or you use Adafactor anyway:

Over at OneTrainer we have optimized the vram of Prodigy a little bit. It's now basically at Adafactor levels. Might be interesting for kohya too (I don't have contact).

masood•11/1/24, 12:19 PM

and can anyone suggest me, what is the best model for realistic human pictures generation in flux dev ?

Zhigarev•11/3/24, 7:50 PM

Hello! Has anyone worked on "LoRA extracted from Full Fine Tune, then LoRA at different network dims"? I'm looking for someone who can help me with this for a small fee! Please DM me. I already have a prepared dataset for the character Conan (Arnold Schwarzenegger), around 100 images, and my own GPU, an RTX 3090, ready for testing.

ZZhigarev Hello! Has anyone worked on "LoRA extracted from Full Fine Tune, then LoRA at di...

wardensc2•11/4/24, 5:47 AM

Yes i already done it. When extract the highest extractable rank which is 640 the size of lora is similar to fp8 checkpoint or gguf q8 full model

Aleksei Naumov•11/4/24, 9:33 AM

can anyone share best practices for training and generation to achieve maximal likeness of real person and generated images?

uisato•11/4/24, 11:19 AM

New concept videos I made for The Voidz! - https://www.youtube.com/watch?v=pyNTS4oYGFU

YouTubeuisato

Spectral Analysis - [Short Concept Video for @thevoidz_ ]

For more experiments, head over to: https://linktr.ee/uisato

#animation #ai #cinematic

AAleksei Naumov can anyone share best practices for training and generation to achieve maximal l...

Furkan Gözükara SECourses•11/4/24, 11:36 AM

this video is the best one you can find https://youtu.be/FvpWy1x5etM

YouTubeSECourses

FLUX Full Fine-Tuning / DreamBooth Training Master Tutorial for Win...

If you want to train FLUX with maximum possible quality, this is the tutorial looking for. In this comprehensive tutorial, you will learn how to install Kohya GUI and use it to fully Fine-Tune / DreamBooth FLUX model. After that how to use SwarmUI to compare generated checkpoints / models and find the very best one to generate most amazing image...

Uuisato New concept videos I made for The Voidz! - https://www.youtube.com/watch?v=pyNTS...

Furkan Gözükara SECourses•11/4/24, 11:37 AM

video is really high res

FFurkan Gözükara SECourses video is really high res

uisato•11/4/24, 11:37 AM

Thank you, Dr. Furkan!

Xcorpion•11/4/24, 12:58 PM

currently trying to train a character lora (Anime, Pony checkpoint), but it also get that animescreen-cap style(usual anime style), wonder if there is a way to train character without those styles and effects

Dario•11/4/24, 8:24 PM

Hi @Furkan Gözükara SECourses, I saw you commented on Boris Noll's LinkedIn post. I'm curious if you have explored these methods of erasing concepts from the Flux base model, and training it for higher resolution output? I would love to learn more about this workflow (possible also his object and concept LoRA training methods). Thank you.

Dario•11/4/24, 8:58 PM

What's the resource you're referring to for style training the 3D-Cartoon-Style-FLUX on Hugging Face?

DDario Hi <@205854764540362752>, I saw you commented on Boris Noll's LinkedIn post. I'm...

Furkan Gözükara SECourses•11/4/24, 9:13 PM

i didnt

Furkan Gözükara SECourses•11/4/24, 9:13 PM

i think that post is not totally honest :d

FFurkan Gözükara SECourses i think that post is not totally honest :d

Dario•11/4/24, 9:14 PM

Thanks for your reply. Hmm really, the base model part or regarding the loRA's? Would you mind to ellaborate?

DDario Thanks for your reply. Hmm really, the base model part or regarding the loRA's? ...

Furkan Gözükara SECourses•11/4/24, 9:17 PM

the erasing concepts doesnt make sense

Furkan Gözükara SECourses•11/4/24, 9:17 PM

and i dont know how he made such script himself

Furkan Gözükara SECourses•11/4/24, 9:17 PM

it is so hard to make such scripts accurately

FFurkan Gözükara SECourses it is so hard to make such scripts accurately

Dario•11/4/24, 9:19 PM

Okay I see. Curious if he'll ever share that script then, What about finetuning the base model to output a higher resolution (2560x2560px)?

DDario Okay I see. Curious if he'll ever share that script then, What about finetuning ...

Furkan Gözükara SECourses•11/4/24, 9:20 PM

well i had tested on myself limitedly and i didnt notice improvements

Furkan Gözükara SECourses•11/4/24, 9:20 PM

but when training a car i dont know

Furkan Gözükara SECourses•11/4/24, 9:20 PM

it also sounds unreal it is over 5 mega pixel

Furkan Gözükara SECourses•11/4/24, 9:20 PM

model is 1 mega pixel

FFurkan Gözükara SECourses well i had tested on myself limitedly and i didnt notice improvements

Dario•11/4/24, 9:24 PM

Thanks for your insight. I was also very suprised about this post since I have never seen something like that before.
Regarding the concept LoRA's I assume he trained 3 style loras for additional "aumotive" looks?

DDario Thanks for your insight. I was also very suprised about this post since I have n...

Furkan Gözükara SECourses•11/4/24, 9:26 PM

i am not sure

FFurkan Gözükara SECourses i am not sure

Dario•11/4/24, 9:29 PM

Ok thank you. Will you dive into more product/object specific (lora) training experiment/tutorials in the future you recon?

DDario Ok thank you. Will you dive into more product/object specific (lora) training ex...

Furkan Gözükara SECourses•11/4/24, 9:31 PM

i may make but too many jobs waiting i need to research sd 3.5

Furkan Gözükara SECourses•11/4/24, 9:32 PM

currently testing model varients impact on lora

Furkan Gözükara SECourses•11/4/24, 9:32 PM

and lora extraction

Furkan Gözükara SECourses•11/4/24, 9:32 PM

to post on civitiai

FFurkan Gözükara SECourses i may make but too many jobs waiting i need to research sd 3.5 😄

Dario•11/4/24, 9:33 PM

Haha makes sense! We'll keep following your work closely, thanks!

theJindesan•11/5/24, 1:14 AM

@Furkan Gözükara SECourses Is there any reason to use kohya over OneTrainer quality-wise for Flux? Is it easy to convert your best koyha config settings to OneTrainer? Also noticed onetrainer supports SD3.5 There are some nice Quality of life things in Onetrainer, like dynamic save anything ondemand, resume from backup, changing samples during training, sample on demand etc. thoughts?

TtheJindesan <@205854764540362752> Is there any reason to use kohya over OneTrainer quality-w...

wardensc2•11/5/24, 5:19 AM

Kohya have most of your thing but you need to learn how to do it. I use Kohya since SD 1.5 never feel that I need to use OneTrainer. Koyha maybe a bit more technical but when you understand it, it's very versatile

Wwardensc2 Kohya have most of your thing but you need to learn how to do it. I use Kohya si...

theJindesan•11/5/24, 5:21 AM

I like Kohya, it just needs some of the quality of life things. like save now.. easy resume, change sample prompts. So many times i want to use GPU for something, but have to wait 20 epochs

would be nice to have a bit more control..

TtheJindesan <@205854764540362752> Is there any reason to use kohya over OneTrainer quality-w...

Furkan Gözükara SECourses•11/5/24, 8:16 AM

Kohya also supports 3.5 . also you can convert my configs

Furkan Gözükara SECourses•11/5/24, 8:16 AM

OneTrainer was lacking previously

Furkan Gözükara SECourses•11/5/24, 8:16 AM

hopefully I will compare and publish configs

TtheJindesan I like Kohya, it just needs some of the quality of life things. like save now.. ...

wardensc2•11/5/24, 12:23 PM

but I need to study again for someone who already know SD for a year or more, people tend to become more lazy especially when Onetrainer does not have something extraordinary than Kohya and also very few people like @Furkan Gözükara SECourses who dedicate time to study Kohya. Finding the right config for lora training or fine tuning is really time consumption and require deep knowledge about AI programming

Wwardensc2 but I need to study again for someone who already know SD for a year or more, pe...

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

you can clone our kohya configs to onetrainer i did it for sdxl and sd 1.5 :d

wardensc2•11/5/24, 12:34 PM

really, how about training speed Kohya vs Onetrainer

Wwardensc2 really, how about training speed Kohya vs Onetrainer

Furkan Gözükara SECourses•11/5/24, 12:37 PM

i havent compared yet but kohya also just made a branch to speed up fine tuning

Wwardensc2 Yes i already done it. When extract the highest extractable rank which is 640 th...

Zhigarev•11/5/24, 1:06 PM

Could you get in touch with me so we can discuss this?

FFurkan Gözükara SECourses i havent compared yet but kohya also just made a branch to speed up fine tuning ...