i published a video of it
i published a video of it
Awhat upscale method ? supir ?
Awhere
FYouTubeSECourses
OmniHuman will shake the AI world forever if get published and works as they claimed. This is literally the ultimate video model announced so far. It is not published yet but I will be following it very closely. I hope that they will publish models publicly so we can use locally hopefully. As soon as published I plan to make APP for this model h...
Fbut no code or model yet
AI think that china will be the next super power
Athey are dropping "bombs" in the next revolutionary field
F100%
Fi hope they bring GPUs
AAlso @Furkan Gözükara SECourses your research has changed my life in a way I want to thank you for everything
AKeep going this community is awesome
AIdk how I didn't find it until September
Fif i missed any question message me again
.
Fnice quality
.@Furkan Gözükara SECourses thank you!
.Feel like I want to train a large fine tune with this method.
.@Furkan Gözükara SECourses thank you for the advice and the amazing configs. Wouldn't have gotten this result without them!
SDr furkan the last model that make china work pretty well woth AMD do you think at least can be models that work good in any GPU brand
.Few more outputs. Still first epoch only.
Fi wish AMD was paying some more attention
Fnice
Fworking on VisoMaster auto installer
Dim trying to use the deepface similiarity app, but i'm getting this error message. Anyone know what this is?
D
Fyes
Fcuda and cudnn installation error
Fplease install cuda 12.4 and cudnn 9.4 or 9.7
Fnice the project working
Fhopefully will publish tomorrow after testing on massed compute too
DI have 12.8 installed. Downgrade?
Fyep dont use cuda 12.8 yet
Fuse 12.4 atm
Vim trying to get caught up on stuff
Vare illustrious models the same as sdxl models?
FHi everyone, and hello @Dr. Furkan Gözükara . Thank you so much for all the work and research you share, it's amazing!
Can I ask your advice on what's the best way to train a LORA for SDXL on cloud as of today (best quality possible)?
Can I ask your advice on what's the best way to train a LORA for SDXL on cloud as of today (best quality possible)?
Fwhich illustrious models?
Fyou can use both Kohya and OneTrainer
Fi recommend onetrainer on massed compute
Fthis is good
F
NI am curious, what exactly are you training here? Such results are already possible witrh Flux (de-distilled) out of the box, right?
.No version is fine tuned this way.
.YouTubePronto CG Tutor
A demo of an image generation of my custom Flux dev de-distilled fine tune trained on high resolution images.
.Ever had the issue when you upscale too high with sd ultimate upscale that your skin details get washed out and lose detail? With this way of training it doesn’t
Nyes, I see. It makes sense! May I ask how are you training it: Are you using custom prompts like ohwx? how does such fine-tuning work, like style training? With every database image captioned in details or not at all? I am trying to understand how the AI model "thinks" in such situations. With custom faces or objects it's straight forward, but with whole "concepts" like this?
MDoes anyone know this error when generating model in kohya
MDoes anyone know this error when generating model in kohya
raceback (most recent call last):
File "<frozen runpy>", line 198, in _run_module_as_main
File "<frozen runpy>", line 88, in _run_code
File "C:\Kohya\kohya_ss\venv\Scripts\accelerate.EXE__main.py", line 4, in <module>
File "C:\Kohya\kohya_ss\venv\Lib\site-packages\accelerate\commands\accelerate_cli.py", line 19, in <module>
from accelerate.commands.estimate import estimate_command_parser
File "C:\Kohya\kohya_ss\venv\Lib\site-packages\accelerate\commands\estimate.py", line 34, in <module>
import timm
File "C:\Kohya\kohya_ss\venv\Lib\site-packages\timm__init.py", line 2, in <module>
from .models import create_model, list_models, is_model, list_modules, model_entrypoint, \
File "C:\Kohya\kohya_ss\venv\Lib\site-packages\timm\models__init__.py", line 28, in <module>
from .maxxvit import *
File "C:\Kohya\kohya_ss\venv\Lib\site-packages\timm\models\maxxvit.py", line 225, in <module>
@dataclass
^^^^^^^^^
File "C:\Python311\Lib\dataclasses.py", line 1232, in dataclass
return wrap(cls)
^^^^^^^^^
File "C:\Python311\Lib\dataclasses.py", line 1222, in wrap
return _process_class(cls, init, repr, eq, order, unsafe_hash,
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Python311\Lib\dataclasses.py", line 958, in _process_class
cls_fields.append(_get_field(cls, name, type, kw_only))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Python311\Lib\dataclasses.py", line 815, in _get_field
raise ValueError(f'mutable default {type(f.default)} for field '
ValueError: mutable default <class 'timm.models.maxxvit.MaxxVitConvCfg'> for field conv_cfg is not allowed: use default_factory
22:42:12-093630 INFO Training has ended.
raceback (most recent call last):
File "<frozen runpy>", line 198, in _run_module_as_main
File "<frozen runpy>", line 88, in _run_code
File "C:\Kohya\kohya_ss\venv\Scripts\accelerate.EXE__main.py", line 4, in <module>
File "C:\Kohya\kohya_ss\venv\Lib\site-packages\accelerate\commands\accelerate_cli.py", line 19, in <module>
from accelerate.commands.estimate import estimate_command_parser
File "C:\Kohya\kohya_ss\venv\Lib\site-packages\accelerate\commands\estimate.py", line 34, in <module>
import timm
File "C:\Kohya\kohya_ss\venv\Lib\site-packages\timm__init.py", line 2, in <module>
from .models import create_model, list_models, is_model, list_modules, model_entrypoint, \
File "C:\Kohya\kohya_ss\venv\Lib\site-packages\timm\models__init__.py", line 28, in <module>
from .maxxvit import *
File "C:\Kohya\kohya_ss\venv\Lib\site-packages\timm\models\maxxvit.py", line 225, in <module>
@dataclass
^^^^^^^^^
File "C:\Python311\Lib\dataclasses.py", line 1232, in dataclass
return wrap(cls)
^^^^^^^^^
File "C:\Python311\Lib\dataclasses.py", line 1222, in wrap
return _process_class(cls, init, repr, eq, order, unsafe_hash,
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Python311\Lib\dataclasses.py", line 958, in _process_class
cls_fields.append(_get_field(cls, name, type, kw_only))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Python311\Lib\dataclasses.py", line 815, in _get_field
raise ValueError(f'mutable default {type(f.default)} for field '
ValueError: mutable default <class 'timm.models.maxxvit.MaxxVitConvCfg'> for field conv_cfg is not allowed: use default_factory
22:42:12-093630 INFO Training has ended.
Fuse Python 3.10 and reinstall
MI sent you a message with the printout I have Python 10..
but its automatic installer uses python 11 alone, which already comes with it
but its automatic installer uses python 11 alone, which already comes with it