Software Engineering Courses (SECourses)•17mo ago

but flux lora cannot be used on kaggle 😦

but flux lora cannot be used on kaggle

Vvirginia but flux lora cannot be used on kaggle 😦

Furkan Gözükara SECourses•8/22/24, 1:22 PM

yep

Furkan Gözükara SECourses•8/22/24, 1:22 PM

forget kaggle

training requires bf16

Furkan Gözükara SECourses•8/22/24, 1:22 PM

full grids published : https://www.reddit.com/r/StableDiffusion/comments/1eyj4b8/kohya_ss_gui_very_easy_flux_lora_trainings_full/

From the StableDiffusion community on Reddit: Kohya SS GUI very eas...

Explore this post and more from the StableDiffusion community

TheToday•8/22/24, 2:05 PM

If I want to make a lora in MC now, what should I do? Update kohya and with what configuration? Is this the Best_v1_5e_5_max_grad_norm_0.json ?

TTheToday If I want to make a lora in MC now, what should I do? Update kohya and with what...

Furkan Gözükara SECourses•8/22/24, 2:11 PM

use latest config in zip file 48_GB_GPUs.json

Furkan Gözükara SECourses•8/22/24, 2:11 PM

you can just use but it will run slower

Furkan Gözükara SECourses•8/22/24, 2:11 PM

unless you manually upgrade torch and prevent auto update

Furkan Gözükara SECourses•8/22/24, 2:11 PM

manually update torch and replace gui.sh file

message_9.txt4.22KB

FFurkan Gözükara SECourses use latest config in zip file 48_GB_GPUs.json

TheToday•8/22/24, 2:18 PM

Ok, I understand that it is the one found in Kohya_GUI_Flux_Installer_v1.zip, but I don't quite understand what I should do with what you sent me

TheToday•8/22/24, 2:24 PM

Could you make a detailed guide with the commands or things to change in MC? Like a massedcompute instructions read haha It would be much appreciated

Experimental_Work•8/22/24, 2:39 PM

How can I apply styles to the entire image in flux? It seems to often generate the subject in the prompt with somewhat boring consistency. Is this a limitation of the model?

TTheToday Ok, I understand that it is the one found in Kohya_GUI_Flux_Installer_v1.zip, bu...

Furkan Gözükara SECourses•8/22/24, 2:50 PM

On Mc you can just run it without doing anything either

Furkan Gözükara SECourses•8/22/24, 2:50 PM

Ah wait you need to do git checkout

Furkan Gözükara SECourses•8/22/24, 2:50 PM

Give me some time to include into script file

EExperimental_Work How can I apply styles to the entire image in flux? It seems to often generate t...

Furkan Gözükara SECourses•8/22/24, 2:51 PM

Currently if you do non realistic style it fails. Pretty overfit. I am trying to fix

TheToday•8/22/24, 3:25 PM

lol I think I put something wrong, it's 200 epoch but in repeats it's 40 which comes by default or should I put 1?

TTheToday lol I think I put something wrong, it's 200 epoch but in repeats it's 40 which c...

Furkan Gözükara SECourses•8/22/24, 3:26 PM

repeat is 1

Furkan Gözükara SECourses•8/22/24, 3:27 PM

we dont use reg images

Furkan Gözükara SECourses•8/22/24, 3:27 PM

you get 8.3 second / it

Furkan Gözükara SECourses•8/22/24, 3:27 PM

it will be 4.4 second after library update

Furkan Gözükara SECourses•8/22/24, 3:27 PM

give me some time to prepare it :d

FFurkan Gözükara SECourses repeat is 1

sergio_gonzo•8/22/24, 3:28 PM

Dr Furkan I hope everything is ok, one question do you ear if onetrainer will support flux?

Ssergio_gonzo Dr Furkan I hope everything is ok, one question do you ear if onetrainer will su...

Furkan Gözükara SECourses•8/22/24, 3:28 PM

not yet. i heard that they are perfecting sd3 first

Furkan Gözükara SECourses•8/22/24, 3:28 PM

but my workflow should workl there

Furkan Gözükara SECourses•8/22/24, 3:28 PM

once they dded

Furkan Gözükara SECourses•8/22/24, 3:29 PM

i have added screenshot of the gui so you can see what to change

FFurkan Gözükara SECourses give me some time to prepare it :d

TheToday•8/22/24, 3:29 PM

Of course thank you

Horrid•8/22/24, 3:59 PM

i need help i get this error when trying to run the kohya train for flux :
Traceback (most recent call last):
File "F:\kohyaflux\Kohya_GUI_Flux_Installer_v1\kohya_ss\sd-scripts\flux_train_network.py", line 411, in <module>
trainer.train(args)
File "F:\kohyaflux\Kohya_GUI_Flux_Installer_v1\kohya_ss\sd-scripts\train_network.py", line 390, in train
self.cache_text_encoder_outputs_if_needed(args, accelerator, unet, vae, text_encoders, train_dataset_group, weight_dtype)
File "F:\kohyaflux\Kohya_GUI_Flux_Installer_v1\kohya_ss\sd-scripts\flux_train_network.py", line 160, in cache_text_encoder_outputs_if_needed
unet.to("cpu")
File "F:\kohyaflux\Kohya_GUI_Flux_Installer_v1\kohya_ss\venv\lib\site-packages\torch\nn\modules\module.py", line 1174, in to
return self._apply(convert)
File "F:\kohyaflux\Kohya_GUI_Flux_Installer_v1\kohya_ss\venv\lib\site-packages\torch\nn\modules\module.py", line 780, in _apply
module._apply(fn)
File "F:\kohyaflux\Kohya_GUI_Flux_Installer_v1\kohya_ss\venv\lib\site-packages\torch\nn\modules\module.py", line 805, in _apply
param_applied = fn(param)
File "F:\kohyaflux\Kohya_GUI_Flux_Installer_v1\kohya_ss\venv\lib\site-packages\torch\nn\modules\module.py", line 1167, in convert
raise NotImplementedError(
NotImplementedError: Cannot copy out of meta tensor; no data! Please use torch.nn.Module.to_empty() instead of torch.nn.Module.to() when moving module from meta to a different device.
Traceback (most recent call last):
File "C:\Users\Surikata\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\Surikata\AppData\Local\Programs\Python\Python310\lib\runpy.py", line 86, in _run_code
exec(code, run_globals)
File "F:\kohyaflux\Kohya_GUI_Flux_Installer_v1\kohya_ss\venv\Scripts\accelerate.EXE__main__.py", line 7, in <module>
File "F:\kohyaflux\Kohya_GUI_Flux_Installer_v1\kohya_ss\venv\lib\site-packages\accelerate\commands\accelerate_cli.py", line 48, in main
args.func(args)
File "F:\kohyaflux\Kohya_GUI_Flux_Installer_v1\kohya_ss\venv\lib\site-packages\accelerate\commands\launch.py", line 1106, in launch_command
simple_launcher(args)
File "F:\kohyaflux\Kohya_GUI_Flux_Installer_v1\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 '['F:\kohyaflux\Kohya_GUI_Flux_Installer_v1\kohya_ss\venv\Scripts\python.exe', 'F:/kohyaflux/Kohya_GUI_Flux_Installer_v1/kohya_ss/sd-scripts/flux_train_network.py', '--config_file', 'F:/AISTUFF/train/model/config_lora-20240822-185636.toml']' returned non-zero exit status 1.
18:56:49-324344 INFO Training has ended.

Horrid•8/22/24, 4:00 PM

i tried using se's config too and it also didnt work im unsure what the issue is

Horrid•8/22/24, 4:02 PM

im downloading the fp16 dev version rn to see if it solv es iy

HHorrid i need help i get this error when trying to run the kohya train for flux : Trac...

Furkan Gözükara SECourses•8/22/24, 4:15 PM

please use dev flux fp16 as base model

Furkan Gözükara SECourses•8/22/24, 4:15 PM

the link in patreon post

Furkan Gözükara SECourses•8/22/24, 4:16 PM

also did you run Fix_For_FLUX_Step_2.bat

Horrid•8/22/24, 4:23 PM

yes i did

HHorrid yes i did

Furkan Gözükara SECourses•8/22/24, 4:28 PM

ok fixed?

Horrid•8/22/24, 4:42 PM

I'll have to test in a bit

mikemenders•8/22/24, 5:15 PM

I see everyone is fluxing. However, I managed to find the golden point for training ponyRealism 2.1 based on the classic SD 1.5 Lora training. So same way with reg images, Prodigy optimizer. What is interesting is that the previous Prodigy parameters hardly needed to be changed, only the d_coef value had to be lowered to 0.4 for 100 images. For about 150 images, it drops to 0.25.
What's important is that I don't use a bucket, I just work with the best 100 images, 1024x1024 resolution, and the classic 'xxxxx woman' folder name. (Where xxxxx is the name of the lora. In Furkan's case ohwx.) Another important thing is that I need 24 GB VRAM here, almost fully exploiting the training. For 100 images it's about 45-50 minutes, for 150 images it's about 1 hour 20 minutes average training.
I just trained 5 Lora, and for all of them the max_norm/keys_scaled value on Tensorboard rarely went above 2, so they became perfect person Lora. I tried it before with 16/16 dim/alpha, but 32/32 turned out to be the right value.
What's interesting is that the image generation got better with euler_a, and bemypony_Photo3 model. However, the Lora performed quite well on the other Pony models, so I'm satisfied. Here's my Kohya json file, feel free to use it!

PonyXL1-realistic-person-sfw.json4.58KB

HHorrid i need help i get this error when trying to run the kohya train for flux : Trac...