Software Engineering Courses (SECourses)•16mo ago

Save Every N Epoch = 1, i thought it would be only 25 versions. but now it will be 100.... Then I s

Save Every N Epoch = 1,
i thought it would be only 25 versions. but now it will be 100.... Then I should keep the best one, right? the one with the best result. That will be the one I will use for my work.

DDEXperience Solutions @Dr. Furkan Gözükara why does it show 52/100 epoch? if I put only 25, what did ...

Maxivi•9/8/24, 5:24 PM

what are you trainign for 10k steps?

FFurkan Gözükara SECourses https://discord.gg/RYYcHwbP

�

𝗡𝗶𝘁𝗶𝗻 𝗣𝗮𝗿𝗲𝘃𝗮•9/8/24, 5:26 PM

Is there a kohya discord server?

DEXperience SolutionsOP•9/8/24, 5:31 PM

800 imgs

DDEXperience Solutions 800 imgs

Maxivi•9/8/24, 5:34 PM

there is an imposter amongnst us

MMaxivi there is an imposter amongnst us

DEXperience SolutionsOP•9/8/24, 5:38 PM

Why do you say that?

�𝗡𝗶𝘁𝗶𝗻 𝗣𝗮𝗿𝗲𝘃𝗮 Is there a kohya discord server?

Furkan Gözükara SECourses•9/8/24, 5:44 PM

Sadly nope

DDEXperience Solutions Save Every N Epoch = 1, i thought it would be only 25 versions. but now it will...

Furkan Gözükara SECourses•9/8/24, 5:45 PM

Yes it will save 100

Xxlads okay i made the unet file inside of models then restarted/updated but still gett...

Sipriyani•9/8/24, 6:00 PM

for my case
I had to remove " --fast " option. i have, desktop and laptop both 4090 , and same problem.

SSipriyani for my case I had to remove " --fast " option. i have, desktop and laptop bo...

xlads•9/8/24, 6:08 PM

i tried this and still getting same issue

SSipriyani for my case I had to remove " --fast " option. i have, desktop and laptop bo...

xlads•9/8/24, 6:15 PM

did you have unet file inside of models?

Aldo Raine•9/8/24, 6:46 PM

File "C:\Users\chris\AppData\Local\Programs\Python\Python310\lib\runpy.py", line 196, in _run_module_as_main
return _run_code(code, main_globals, None,
File "C:\Users\chris\AppData\Local\Programs\Python\Python310\lib\runpy.py", line 86, in _run_code
exec(code, run_globals)
File "D:\Kohya_GUI_Flux_Installer_30\kohya_ss\venv\Scripts\accelerate.EXE__main__.py", line 7, in <module>
File "D:\Kohya_GUI_Flux_Installer_30\kohya_ss\venv\lib\site-packages\accelerate\commands\accelerate_cli.py", line 48, in main
args.func(args)
File "D:\Kohya_GUI_Flux_Installer_30\kohya_ss\venv\lib\site-packages\accelerate\commands\launch.py", line 1106, in launch_command
simple_launcher(args)
File "D:\Kohya_GUI_Flux_Installer_30\kohya_ss\venv\lib\site-packages\accelerate\commands\launch.py", line 704, in simple_launcher
raise subprocess.CalledProcessError(returncode=process.returncode, cmd=cmd)
subprocess.CalledProcessError: Command '['D:\Kohya_GUI_Flux_Installer_30\kohya_ss\venv\Scripts\python.exe', 'D:/Kohya_GUI_Flux_Installer_30/kohya_ss/sd-scripts/flux_train_network.py', '--config_file', '/home/Ubuntu/apps/StableSwarmUI/Models/Lora/config_lora-20240908-123856.toml']' returned non-zero exit status 3221225477.
12:40:57-496469 INFO Training has ended.

Aldo Raine•9/8/24, 6:46 PM

got the above when I started my training

JohnDopamine•9/8/24, 7:07 PM

When T5 / clip-l testing commencing? Need info on what to use, LR to use, etc

JJohnDopamine When T5 / clip-l testing commencing? Need info on what to use, LR to use, etc

Maxivi•9/8/24, 7:11 PM

t5 very sensetive, kohya is implementing different learning rates soon for each

Maxivi•9/8/24, 7:11 PM

so save your time and energy and wait for it

JohnDopamine•9/8/24, 7:25 PM

I'm using a beta of a friends trainer that has split LR - but honestly from my trials so far they work pretty well linked.

Regardless, this is why I'd like to see thorough systematic "research"

JohnDopamine•9/8/24, 7:26 PM

What LRs are you using w/ Kohyas at the moment?

�𝗡𝗶𝘁𝗶𝗻 𝗣𝗮𝗿𝗲𝘃𝗮 Is there a kohya discord server?

GomezBro•9/8/24, 7:31 PM

Your on it now right? Lol this is it.

JJohnDopamine What LRs are you using w/ Kohyas at the moment?

Maxivi•9/8/24, 7:42 PM

kohya told me 1/10th the base unet lr

Maxivi•9/8/24, 7:42 PM

so if you're doing 5e-05 for unet do 5e-06 for te

Merc•9/8/24, 7:43 PM

yo what is that

Merc•9/8/24, 7:43 PM

im so cooked

FFurkan Gözükara SECourses reply to guy who replied you

xlads•9/8/24, 7:49 PM

my flux model is now working inside of swarmui, i had to reintall swarmui and select "flex" in the step to install swarmui

xlads•9/8/24, 8:31 PM

Phuong Bui•9/8/24, 8:33 PM

Is there currently any method to train lora for small details like that? I tried training but when I use it, there are almost no such small details or if there are, the text will not be the same. Anyone who has trained has any experience to share?

DEXperience SolutionsOP•9/8/24, 9:27 PM

is it possible to train a model or a lora, to make this kind of comic images?

SSipriyani for my case I had to remove " --fast " option. i have, desktop and laptop bo...

Furkan Gözükara SECourses•9/8/24, 9:32 PM

this happens when comfyui torch is not up to date

AAldo Raine File "C:\Users\chris\AppData\Local\Programs\Python\Python310\lib\runpy.py", line...

Furkan Gözükara SECourses•9/8/24, 9:32 PM

this doesnt tell the error

JJohnDopamine When T5 / clip-l testing commencing? Need info on what to use, LR to use, etc

Furkan Gözükara SECourses•9/8/24, 9:33 PM

i may start today good reminding

Xxlads my flux model is now working inside of swarmui, i had to reintall swarmui and se...

Furkan Gözükara SECourses•9/8/24, 9:33 PM

nice

Xxlads Click to see attachment

Furkan Gözükara SECourses•9/8/24, 9:33 PM

nice one

PPhuong Bui Is there currently any method to train lora for small details like that? I tried...

Furkan Gözükara SECourses•9/8/24, 9:34 PM

nope i havent tried yet

DDEXperience Solutions is it possible to train a model or a lora, to make this kind of comic images?

Furkan Gözükara SECourses•9/8/24, 9:35 PM

can be doable with a good dataset

FFurkan Gözükara SECourses nope i havent tried yet

Phuong Bui•9/8/24, 10:07 PM

If you have time, can you try it? Thank you very much

PPhuong Bui If you have time, can you try it? Thank you very much

Furkan Gözükara SECourses•9/8/24, 10:33 PM

currently super busy but i plan in future product training

Hustlers4Ever•9/8/24, 11:35 PM

did something update recently? entry not found in URL: https://huggingface.co/Kijai/flux-fp8/resolve/main/flux1-dev-fp8.safetensors. getting this error when installing FLUX.1-dev FP8 Version model on runpod, no issues yesterday

HHustlers4Ever did something update recently? entry not found in URL: https://huggingface.co/Ki...

Furkan Gözükara SECourses•9/8/24, 11:38 PM

yep deleted

Furkan Gözükara SECourses•9/8/24, 11:38 PM

extremely stupid move

Furkan Gözükara SECourses•9/8/24, 11:38 PM

https://huggingface.co/Kijai/flux-fp8/tree/main

Kijai/flux-fp8 at main

Furkan Gözükara SECourses•9/8/24, 11:39 PM

i dont know which one better but probably e4m3fn.safetensors

Furkan Gözükara SECourses•9/8/24, 11:40 PM

The flux1-dev-fp8 was just renamed to flux1-dev-fp8-e4m3fn to better reflect the format of the weights, nothing in the file has changed, no need to re-download that.

Furkan Gözükara SECourses•9/8/24, 11:41 PM

probably i was using that link as well now i have to change all

Furkan Gözükara SECourses•9/8/24, 11:41 PM

i told him to reupload i hope he does

FFurkan Gözükara SECourses extremely stupid move

Hustlers4Ever•9/8/24, 11:49 PM

sounds good thank you

Furkan Gözükara SECourses•9/8/24, 11:51 PM

2x upscale add those noise