Software Engineering Courses (SECourses)•3y ago

Hi guys, in Dreambooth, I set the 'Class Images Per Instance Image' to 50 and added 1500 images to t

Hi guys, in Dreambooth, I set the 'Class Images Per Instance Image' to 50 and added 1500 images to the 'Classification Dataset Directory.' I checked the folder path and it is correct. However, SD is still attempting to generate 1500 images. Should I set 'Class Images Per Instance Image' to 0?
Thank you very much.

KKhang Hi guys, in Dreambooth, I set the 'Class Images Per Instance Image' to 50 and ad...

Furkan Gözükara SECourses•4/27/23, 4:14 PM

show me screenshot of path please

Daniel•4/27/23, 4:14 PM

Hi, i have question about SD Dreambooth: have they fixed the problems with it? i've read about it getting broken after some updated and not producing good results. on my first attempts with it (using lora)i didnt have great results, following you instructions. although my data set only had me in different angle and not distances and cloths. wanted to know if the tool is the problem or i just need to try more stuff.

DDaniel Hi, i have question about SD Dreambooth: have they fixed the problems with it? i...

Furkan Gözükara SECourses•4/27/23, 4:14 PM

currently it is working very well

Furkan Gözükara SECourses•4/27/23, 4:14 PM

they fixed bugs

Daniel•4/27/23, 4:17 PM

okay nice, and i have a 3060 ti with 8gb vram. do i need to use lora or i can run it without it?

DDaniel okay nice, and i have a 3060 ti with 8gb vram. do i need to use lora or i can ru...

Furkan Gözükara SECourses•4/27/23, 4:18 PM

you need lora

Furkan Gözükara SECourses•4/27/23, 4:18 PM

you cant do dreambooth

Furkan Gözükara SECourses•4/27/23, 4:18 PM

actually right now i am post processing kohya lora tutorial

FFurkan Gözükara SECourses actually right now i am post processing kohya lora tutorial

Daniel•4/27/23, 4:19 PM

i will catch that

Daniel•4/27/23, 4:19 PM

thank you

xariis_f•4/27/23, 5:05 PM

Hi everyone, I hope you're all doing great. I followed this tutorial https://youtu.be/dYt9xJ7dnpU to install Kandinsky 2.1 and generate AI art using my graphics card. I followed every step and it was going well until I ran this code to download the files and models:
model = get_kandinsky2(
'cuda',
task_type='text2img',
cache_dir='/tmp/kandinsky2',
model_version='2.1',
use_flash_attention=False
)
I tried deleting and re-downloading the files, but the same error keeps appearing every time. If anyone has any ideas on how I can fix this, it would be greatly appreciated. Thank you.

YouTubeSECourses

Midjourney Level NEW Open Source Kandinsky 2.1 Beats Stable Diffusi...

Discord : https://bit.ly/SECoursesDiscord. Kandinsky 2.1 is truly exceptional, and it is on par with Midjourney. In this video, I will compare Kandinsky to Stable Diffusion and provide a comprehensive tutorial on installation and usage. If I have been of assistance to you and you would like to show your support for my work, please consider becom...

xariis_f•4/27/23, 5:06 PM

Hiccup•4/27/23, 5:34 PM

Is your python path set correctly xD

Crazotor•4/27/23, 6:21 PM

hi guys, I have just started to learn SD for a week now. I am trying to train with dreambooth but I just cant get it to work like I want to. I have seen SEcourses video's for the best settings and tried those but I am a little confused which ones to use. I have a RTX a4000 card with 16GB of vram. Should i use EMA and 8bit adam or lion, which mixed precision and somehow i cant use xformers (I can only use the default option). Can you guys help me?

Xxariis_f Hi everyone, I hope you're all doing great. I followed this tutorial https://you...

Furkan Gözükara SECourses•4/27/23, 6:47 PM

you can download manually

CCrazotor hi guys, I have just started to learn SD for a week now. I am trying to train wi...

Furkan Gözükara SECourses•4/27/23, 6:48 PM

yes ema and lion best

Furkan Gözükara SECourses•4/27/23, 6:48 PM

you can use xformers must be related to your installation

Furkan Gözükara SECourses•4/27/23, 6:48 PM

i have a very interesting idea working on that right now. using AI to sort most similar generated images to quickly find best ones :d

Furkan Gözükara SECourses•4/27/23, 6:49 PM

generated over 1000 images so with AI i will be able to quickly get good ones :d

Zono50•4/27/23, 7:03 PM

hey doc, how do i use the tokenizer extension to see how much weight my prompts are carrying?

Zono50•4/27/23, 7:03 PM

i'm using a lora model and doesn't seem like it's really influencing this model with my current prompt

Zono50•4/27/23, 7:04 PM

i also tried daam but getting weird errors

Zono50•4/27/23, 7:19 PM

i watched your tutorial, when i put my lora tag in daam, it doesn't show any heat at all, is that normal?

ZZono50 i watched your tutorial, when i put my lora tag in daam, it doesn't show any hea...

Furkan Gözükara SECourses•4/27/23, 7:40 PM

i think it doesnt support lora

Furkan Gözükara SECourses•4/27/23, 7:40 PM

you need full ckpt

Furkan Gözükara SECourses•4/27/23, 7:54 PM

dmn this thing working great. i am sorting 5000 generated images to find most similar to me :d

Furkan Gözükara SECourses•4/27/23, 7:55 PM

@xariis_f do you have any questions?

Zono50•4/27/23, 7:55 PM

looks like it does

Zono50•4/27/23, 8:01 PM

although that's not my model lol

Furkan Gözükara SECourses•4/27/23, 9:12 PM

according to AI

Furkan Gözükara SECourses•4/27/23, 9:12 PM

this is the most similar image

Furkan Gözükara SECourses•4/27/23, 9:12 PM

Furkan Gözükara SECourses•4/27/23, 9:13 PM

to this one

Furkan Gözükara SECourses•4/27/23, 9:13 PM

what you think

Furkan Gözükara SECourses•4/27/23, 9:14 PM

it passes 5 of 6 ai face recognition

rimasach•4/27/23, 9:25 PM

I'm stuck here. trying to use control net + stable diffuser

Rrimasach I'm stuck here. trying to use control net + stable diffuser

Furkan Gözükara SECourses•4/27/23, 9:51 PM

installation error

rimasach•4/27/23, 9:52 PM

how to fix it

Furkan Gözükara SECourses•4/27/23, 9:52 PM

follow this tutorial please : https://www.youtube.com/watch?v=tXaQAkOgezQ

YouTubeSECourses

New Style Transfer Extension, ControlNet of Automatic1111 Stable Di...

Discord : https://bit.ly/SECoursesDiscord. New fantastic style transfer feature via T2I-Adapter added to the #ControlNet extension. 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