How far did you train this lora ? Like the steps and epochs
How far did you train this lora ? Like the steps and epochs
YNeither lol
YI used very low quality photos, 20 photos / 200 epochs with high quality photos this could be much better
YI think i found a way to minimize artifacts too
SHow so?
SYou mentioned that you tweaked the config a little bit. Do you mind sharing that?
YIn the config i use another scheduler polynomial scheduler
YTo reduce overfitting at the end of training
Ybut this only for training
Yin my comfyui i coded 2 nodes that helped improve generation quality
YFor artifacts
YI do multiple sampling
Ywith many resolutions
SCan this be found right inside kohya ?
YI don't know but i changed in config.toml
YI directly use sd scripts
YI don't use gui
SHow exactly does this reduce artifacts?
SOhh now I see
YI will be sharing everything once i complete my workflow bro
Fmore better but there is no such min number
FPatreon
Get more from SECourses: FLUX, Tutorials, Guides, Resources, Training, Scripts on Patreon
YI am trying now many tests now to find best possible way for generation
Y@Dr. Furkan Gözükara Saw sdxl lastest release
Y?
Y3 models
Yblur canny i guess and another one
Y@Sebastianblood
Ylooks good ?
VI guess I cannot finetune flux there? 
YThis is with artifacts :
YThis is optimized see jewelery :
VWas not able to watch all parts of the video for fine tuning flux yet. only massed compute and some parts of other previous tutorial video.
I didnt see yet the part where is discussed quality difference between 15 images and 250 images for training. I am curious how big or in what range is the difference in quality?
I didnt see yet the part where is discussed quality difference between 15 images and 250 images for training. I am curious how big or in what range is the difference in quality?
A@Dr. Furkan Gözükara where can I find what every new flux model does ?
T
YThey say it has been leaked
YOpenAI sora
Ycheck code source turned out they use api but maybe didnt dig enough
Y
YThey gave artists access to check it out
Ysomeone took the api access and put it to the public
Ylol
TxD
V@Dr. Furkan Gözükara ah sorry if previous messages sounded rude or sarcastic. I realized they might appear in such a way. But rest assured, if I ask such questions, its because I never used swarmui or kaggle before. I would say first one is a UI and the second term I dont know yeah I guess I need to watch the videos 
yeah I guess I need to watch the videos Va different question: I heard in the news there are problems with scaling or with training new models. Does somebody know more about this?
Fyou cant find you have to search internet follow repos
Fread patreon post there are more grids
Fopen ai scrwd them
Fno only api
Sface skin is not as vivid as the other pictures you shared earlier
GAre you saying they only had the API? So no source code was leaked?