Software Engineering Courses (SECourses)•17mo ago

I trained flux TE 1e04 and unet 1e04 with pictures of me and a friend after 15 epochs all man looked

I trained flux TE 1e04 and unet 1e04 with pictures of me and a friend after 15 epochs all man looked like me on the samples, I think I'm overtraining with too high learning rate, now I'm testing with 5e05 unet and 1e04 TE I leave the TE high to see its effects on the training.

Furkan Gözükara SECourses•8/28/24, 2:08 AM

amazing news

Furkan Gözükara SECourses•8/28/24, 2:08 AM

torch 2.4.1 gonna fix slowness on windows

Ddsienra I trained flux TE 1e04 and unet 1e04 with pictures of me and a friend after 15 e...

Furkan Gözükara SECourses•8/28/24, 2:08 AM

good way

Furkan Gözükara SECourses•8/28/24, 2:08 AM

my trainings running

Furkan Gözükara SECourses•8/28/24, 2:08 AM

i will update config file hopefully once i find best spot

dsienraOP•8/28/24, 2:11 AM

may be with text encoder regularization images can be used

Ddsienra may be with text encoder regularization images can be used

Furkan Gözükara SECourses•8/28/24, 2:15 AM

yep need to be testedf

dsienraOP•8/28/24, 2:15 AM

the speed es much better now, I restarted the pc I'm with WSL rtx3090

Ddsienra the speed es much better now, I restarted the pc I'm with WSL rtx3090

Furkan Gözükara SECourses•8/28/24, 2:15 AM

amazing

Furkan Gözükara SECourses•8/28/24, 2:15 AM

like 2x of me on windows

Furkan Gözükara SECourses•8/28/24, 2:15 AM

i hope torch 2.4.1 comes asap

dsienraOP•8/28/24, 2:17 AM

WSL was not so hard to set up

Furkan Gözükara SECourses•8/28/24, 2:18 AM

generating prompts for showcase

Ddsienra WSL was not so hard to set up

Furkan Gözükara SECourses•8/28/24, 2:18 AM

it is hard for many

dsienraOP•8/28/24, 2:19 AM

wsl -d ubuntu

sudo apt install python3.10-venv

wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64/cuda-keyring_1.0-1_all.deb

sudo dpkg -i cuda-keyring_1.0-1_all.deb

sudo apt-key del 7fa2af80

wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64/cuda-keyring_1.0-1_all.deb

sudo dpkg -i cuda-keyring_1.0-1_all.deb

sudo apt-get update

wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64/cuda-keyring_1.0-1_all.deb

sudo apt-get -y install cuda

git clone --recursive https://github.com/bmaltais/kohya_ss.git

chmod +x ./setup.sh

sudo apt update -y

sudo apt install -y python3-tk

sudo apt install software-properties-common -y

sudo add-apt-repository ppa:deadsnakes/ppa

sudo apt update

sudo apt install python3.10 python3.10-venv python3.10-dev

python3 --version

./gui.sh

git branch

git switch sd3-flux.1

source venv/bin/activate

pip install torch==2.4.0+cu124 --index-url https://download.pytorch.org/whl/cu124

pip install torchvision==0.19.0+cu124 --index-url https://download.pytorch.org/whl/cu124

pip install xformers==0.0.27.post2

./gui.sh

GitHub

GitHub - bmaltais/kohya_ss

Contribute to bmaltais/kohya_ss development by creating an account on GitHub.

Ddsienra wsl -d ubuntu sudo apt install python3.10-venv wget https://developer.download...

Furkan Gözükara SECourses•8/28/24, 2:19 AM

code format it

Furkan Gözükara SECourses•8/28/24, 2:20 AM

also how do you handle file saving setup?

dsienraOP•8/28/24, 2:21 AM

that is the list of commands I used to set it up

dsienraOP•8/28/24, 2:21 AM

I save it on a txt

Furkan Gözükara SECourses•8/28/24, 2:21 AM

it probably cant reach your windows folders

dsienraOP•8/28/24, 2:21 AM

yes it can

Furkan Gözükara SECourses•8/28/24, 2:21 AM

i see so its aim is not isolation

dsienraOP•8/28/24, 2:22 AM

Furkan Gözükara SECourses•8/28/24, 2:22 AM

i see what you mean

dsienraOP•8/28/24, 2:22 AM

Furkan Gözükara SECourses•8/28/24, 2:22 AM

i think i may make seperate tutorial for it

dsienraOP•8/28/24, 2:24 AM

by default the windows drives are mounted under \wsl.localhost\Ubuntu\mnt

dsienraOP•8/28/24, 2:27 AM

these are my paths on kohya

dsienraOP•8/28/24, 2:27 AM

im using my windows folders

Ddsienra these are my paths on kohya

Furkan Gözükara SECourses•8/28/24, 2:29 AM

i see thanks for info

FFurkan Gözükara SECourses i see thanks for info

dsienraOP•8/28/24, 2:30 AM

you welcome

Ddsienra wsl -d ubuntu sudo apt install python3.10-venv wget https://developer.download...

nicearoni•8/28/24, 2:44 AM

this is literally what i've just been trying to do for the last hour lol - thank you so much!! i got the kohya gui to launch, but am getting cuda errors, so i haven't bothered trying since that's the whole point over using native windows lol. i may try, though, wondering how soon torch fix is coming?

dsienraOP•8/28/24, 2:52 AM

wsl -d ubuntu

dsienraOP•8/28/24, 2:52 AM

this command is to enter to the one installed distributions

dsienraOP•8/28/24, 2:53 AM

you have to install it first

dsienraOP•8/28/24, 2:56 AM

"wsl --install" first

Nnicearoni this is literally what i've just been trying to do for the last hour lol - thank...

dsienraOP•8/28/24, 3:08 AM

these are the commands I used but is not a tutorial.. some commands only work if you are in the correct path

nicearoni•8/28/24, 3:10 AM

yeah, i know i didn't run the chmod and see some other stuff in there i probably didn't do, or not in the right order so i'll see - tomorrow though, started a big joy caption batch haha

the error i was getting was like the below, plus a couple more similar

Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered

dsienraOP•8/28/24, 3:13 AM

chmod +x ./setup.sh

dsienraOP•8/28/24, 3:13 AM

git switch sd3-flux.1

source venv/bin/activate

pip install torch==2.4.0+cu124 --index-url https://download.pytorch.org/whl/cu124

pip install torchvision==0.19.0+cu124 --index-url https://download.pytorch.org/whl/cu124

pip install xformers==0.0.27.post2

dsienraOP•8/28/24, 3:13 AM

for this you must to be in the kohya folder

FFurkan Gözükara SECourses i see thanks for info

dsienraOP•8/28/24, 4:52 AM

one note, I'm not running kohya from the windows file system, I git cloned kohya to the linux file system Ubuntu\home\user\ I use the windows folders to load the models, images and to save the checkpoints but I don't know if you can run kohya from a mounted ntfs drive so I installed it on the Linux virtual hard drive ext4.vhdx that is on C:\Users<YourUsername>\AppData\Local\Packages<DistroPackageName>\LocalState\ext4.vhdx, this is transparent to the user but is something that is good to know

Charkel•8/28/24, 8:06 AM

Runpod instructions not very clear for SwarmUI. I don't understand what to do

Charkel•8/28/24, 8:16 AM

Also I don't know how much storage I need to set

Nnicearoni this is literally what i've just been trying to do for the last hour lol - thank...

Furkan Gözükara SECourses•8/28/24, 8:51 AM

fix is on the wayt

Furkan Gözükara SECourses•8/28/24, 8:51 AM

more than 75% completed

Furkan Gözükara SECourses•8/28/24, 8:51 AM

i will look to install dev version of torch 2.4.1

CCharkel Runpod instructions not very clear for SwarmUI. I don't understand what to do 😦

Furkan Gözükara SECourses•8/28/24, 8:52 AM

upload all files inside zip file

Furkan Gözükara SECourses•8/28/24, 8:52 AM

set storage like min 100 gb

Furkan Gözükara SECourses•8/28/24, 8:52 AM

and in this video i have shown actually how to install and use : https://youtu.be/XFUZof6Skkw

YouTubeSECourses

How to Use SwarmUI & Stable Diffusion 3 on Cloud Services Kaggle (f...

In this video, I demonstrate how to install and use #SwarmUI on cloud services. If you lack a powerful GPU or wish to harness more GPU power, this video is essential. You'll learn how to install and utilize SwarmUI, one of the most powerful Generative AI interfaces, on Massed Compute, RunPod, and Kaggle (which offers free dual T4 GPU access for ...

I trained flux TE 1e04 and unet 1e04 with pictures of me and a friend after 15 epochs all man looked

Similar Threads

Similar Threads

Similar Threads