Software Engineering Courses (SECourses)•3y ago

@Dr. Furkan Gözükara where can I learn about training with captions? Primarily, trying to figure out

@Dr. Furkan Gözükara where can I learn about training with captions? Primarily, trying to figure out if I need a caption file for each image or a single caption file with a special formatting to associate with each image?

stillmotionOP•10/23/23, 7:28 PM

ah, just looked at the source and it appears to be a single caption file for each image

Iiamannakogan Can someone please point me: I trained kohya SDXL on my photos using the command...

Furkan Gözükara SECourses•10/23/23, 8:41 PM

for console running you need diffusers

Furkan Gözükara SECourses•10/23/23, 8:41 PM

example code in this video : https://youtu.be/kIyqAdd_i10

YouTubeSECourses

Stable Diffusion Google Colab, Continue, Directory, Transfer, Clone...

Our Discord : https://discord.gg/HbqgGaZVmr. This is the video where you will learn how to use Google Colab for Stable Diffusion. 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...

Stable Diffusion Google Colab, Continue, Directory, Transfer, Clone...

Mmikemenders @Dr. Furkan Gözükara check this app, LCM_Dreamshaper on CPU only with OpenVINO t...

Furkan Gözükara SECourses•10/23/23, 8:42 PM

pretty slow

Furkan Gözükara SECourses•10/23/23, 8:42 PM

it takes 1 sec with tensor RT

Furkan Gözükara SECourses•10/23/23, 8:42 PM

,but for those who dont have gpu can work

Sstillmotion @Dr. Furkan Gözükara where can I learn about training with captions? Primarily, ...

Furkan Gözükara SECourses•10/23/23, 8:42 PM

not that i know where. i read papers and did 100s trainings to understand :d

stillmotionOP•10/23/23, 8:44 PM

oh I see what you're saying

Furkan Gözükara SECourses•10/23/23, 8:44 PM

i mean how captions affect training

Furkan Gözükara SECourses•10/23/23, 8:44 PM

you asking that?

stillmotionOP•10/23/23, 8:44 PM

no just how to load them into kohya

stillmotionOP•10/23/23, 8:44 PM

got it now!

Furkan Gözükara SECourses•10/23/23, 8:44 PM

just .txt

Furkan Gözükara SECourses•10/23/23, 8:44 PM

and same file name

Furkan Gözükara SECourses•10/23/23, 8:44 PM

as image file name

Kallamamran•10/23/23, 8:46 PM

My img2img ControlNet is borked

Kallamamran•10/23/23, 8:46 PM

There is no image-area

Kallamamran•10/23/23, 8:47 PM

Ah. I had to check "Upload independent control image"

stillmotionOP•10/23/23, 8:47 PM

@Dr. Furkan Gözükara definitely check out LLaVa for image captioning. So much better than BLIP

Sstillmotion @Dr. Furkan Gözükara definitely check out LLaVa for image captioning. So much be...

Furkan Gözükara SECourses•10/23/23, 8:49 PM

i tested their online version

Furkan Gözükara SECourses•10/23/23, 8:49 PM

waiting them to release windows

Furkan Gözükara SECourses•10/23/23, 8:49 PM

not working on windows yet sadly

stillmotionOP•10/23/23, 8:49 PM

ah no good

Furkan Gözükara SECourses•10/23/23, 8:49 PM

it has checkbox

KKallamamran My img2img ControlNet is borked 😦

Furkan Gözükara SECourses•10/23/23, 8:49 PM

you see there is checkbox

stillmotionOP•10/23/23, 8:50 PM

These guys used LLaVa to caption an entire dataset https://pixart-alpha.github.io/

PIXART-α:Fast Training of Diffusion Transformer for Photorealistic ...

Sstillmotion These guys used LLaVa to caption an entire dataset https://pixart-alpha.github.i...

Furkan Gözükara SECourses•10/23/23, 8:51 PM

ye it is a good idea

Furkan Gözükara SECourses•10/23/23, 8:51 PM

leave a reply here

Furkan Gözükara SECourses•10/23/23, 8:51 PM

https://github.com/haotian-liu/LLaVA/issues/521

GitHub

Can you add venv installer? · Issue #521 · haotian-liu/LLaVA

Describe the issue Conda failing on Windows I have miniconda and anaconda as well (base) PS C:\Users\King> cd "G:\LLaVA" (base) PS G:\LLaVA> conda create -n llava python=3.10 -y Ret...

Furkan Gözükara SECourses•10/23/23, 9:38 PM

rtx 4090 speed : https://twitter.com/GozukaraFurkan/status/1716569763940372822

Kallamamran•10/23/23, 9:42 PM

Seen FreeU? https://github.com/ChenyangSi/FreeU

GitHub

GitHub - ChenyangSi/FreeU: FreeU: Free Lunch in Diffusion U-Net

FreeU: Free Lunch in Diffusion U-Net. Contribute to ChenyangSi/FreeU development by creating an account on GitHub.

KKallamamran Seen FreeU? https://github.com/ChenyangSi/FreeU

Furkan Gözükara SECourses•10/23/23, 9:44 PM

yes only difference is their prompting

Furkan Gözükara SECourses•10/23/23, 9:44 PM

ah this is not that

Furkan Gözükara SECourses•10/23/23, 9:44 PM

ye i tested it

Furkan Gözükara SECourses•10/23/23, 9:45 PM

never gotten better results

Kallamamran•10/23/23, 9:45 PM

Ok. Looked good

Furkan Gözükara SECourses•10/23/23, 9:45 PM

i tested extension

Kallamamran•10/23/23, 9:45 PM

Didin't find one for A1111

Kallamamran•10/23/23, 9:45 PM

Only Comfy

KKallamamran Didin't find one for A1111

Furkan Gözükara SECourses•10/23/23, 9:54 PM

there is

Furkan Gözükara SECourses•10/23/23, 9:55 PM

https://github.com/ljleb/sd-webui-freeu

GitHub

GitHub - ljleb/sd-webui-freeu: a1111 implementation of https://gith...

a1111 implementation of https://github.com/ChenyangSi/FreeU - GitHub - ljleb/sd-webui-freeu: a1111 implementation of https://github.com/ChenyangSi/FreeU

Kallamamran•10/23/23, 9:55 PM

Well I take your word for it that it is not worth the time

Furkan Gözükara SECourses•10/23/23, 9:57 PM

well at least it was for me

Furkan Gözükara SECourses•10/23/23, 9:57 PM

give it a try

iamannakogan•10/23/23, 9:58 PM

How long should a training take on a machine with A10 GPU? 24Gb parameters config with 14 photos -> 4480 iterations says it'll take 2.5 hours - do you think it's OK or too slow?

Iiamannakogan How long should a training take on a machine with A10 GPU? 24Gb parameters confi...

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

it makes 4480 / 150

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

so around 2 second / it

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

a10 is probably slower than rtx 3090 ti

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

so looks accurate

vKorby•10/23/23, 10:50 PM

Hey guys jus joined in here, I've been trying to figure out if this is stable diffusion that they're using? if so I’m really trying to get this type of model because i have no idea what it is, would really appreciate if anyone could help. Thanks

@Dr. Furkan Gözükara where can I learn about training with captions? Primarily, trying to figure out

Similar Threads

Similar Threads

Similar Threads