Dr Furkan in the patreon ther is a link? for that kohya version?
Dr Furkan in the patreon ther is a link? for that kohya version?
Dinteresting, theres some weird degradation in comfyui for flux with continous generation. everything starts ok, but each new generation gets worse till its like this:
RHi @Dr. Furkan Gözükara, I am wondering if you will make a YouTube tutorial on this, or if this will stay in the Patreon. How long will this stay in Patreon only?
https://www.patreon.com/posts/110293257
https://www.patreon.com/posts/110293257
Patreon
Get more from SECourses: Tutorials, Guides, Resources, Training, FLUX, MidJourney, Voice Clone, TTS, ChatGPT, GPT, LLM, Scripts by Furkan Gözü on Patreon
Fi use latest kohya ss
Fsd3 flux branch
Fi didnt determine yet
Fatm patreon only
Fsadly there is no course as far as i know
Fbut i give private lectures
RAlso @Dr. Furkan Gözükara, do you see any difference between LoRA and Full-Fine Tuning for FLUX?
RWhat do you mean by Full-Fine Tuning? Like the training modifies all of FLUX parameters?
RDoes this Patreon link provide a one-click installer that you usually do, or not yet?
https://www.patreon.com/posts/110293257
https://www.patreon.com/posts/110293257
Patreon
Get more from SECourses: Tutorials, Guides, Resources, Training, FLUX, MidJourney, Voice Clone, TTS, ChatGPT, GPT, LLM, Scripts by Furkan Gözü on Patreon
Fi believe fine tuning will be way better
Fi will add fine tuning reserch once lora completed
Fkohya is 1 click install
Fby itself
Flcm makes cgi style
F
Fgonna add samplyer scheduler diff to the post in a moment
ROk LoRA research is in progress, and Full Fine Tuning research is planned?
Fafter lora
RAre you now fully just using swarmUI and no more A1111?
Fcurrently for flux yes
Fthe realism is just great and i am not even cherry picking or best config yet
R@Dr. Furkan Gözükara, how fast is FLUX with your graphics card? Which graphics card is good – is it the one with more VRAM?
I am trying to decide between SDXL Turbo or FLUX...
I am trying to decide between SDXL Turbo or FLUX...
FKohya SS GUI FLUX LoRA experiments - The realism is just great and i am not even cherry picking (these are from my x/y/z grid testings for research) or best config yet (still lots of test to do and evaluate) For stylization it is still super overfit. Trying to fix that as well
F
Fyou need 24 gb gpu atm
For cloud
FPatreon
Get more from SECourses: Tutorials, Guides, Resources, Training, FLUX, MidJourney, Voice Clone, TTS, ChatGPT, GPT, LLM, Scripts by Furkan Gözü on Patreon
Ftesting flux cfg now
Jthis it happened for me too in SwarmUI, the Comfy guys are aware of this problem
Fwow this is a big bug
Fit is certainly a bug
Jyes sure, it happpens if you create alot of pics without stop
DI changed to a different workflow and that fixed it
Fnice
CRegarding style overfitting
what did you tag your photos with? did you tag taken on smartphone or something like amateur photography?
Also have you tried adding back in generated photos of yourself (the ones with good likeness) that dont have that standard style. Or are you only experimenting with how good a model can do if you have very basic dataset
Finally another thing i found to fix style is generating first few steps with 0.4 Lora weight, then generating the remaining with 1.0 weight. Since the first few steps are dictating colors and style it gives it much better style that is closer to what original model would have produced
All the above has helped my loras be much more versatile being able to generate much more natural looking images while retaining likeness
what did you tag your photos with? did you tag taken on smartphone or something like amateur photography?
Also have you tried adding back in generated photos of yourself (the ones with good likeness) that dont have that standard style. Or are you only experimenting with how good a model can do if you have very basic dataset
Finally another thing i found to fix style is generating first few steps with 0.4 Lora weight, then generating the remaining with 1.0 weight. Since the first few steps are dictating colors and style it gives it much better style that is closer to what original model would have produced
All the above has helped my loras be much more versatile being able to generate much more natural looking images while retaining likeness
Fonly ohwx man atrm
Fhow do you do that=? i am gonna ask swarmui :"steps with 0.4 Lora weight, then generating the remaining with 1.0 weight"
JDr.Furkan i can not find the menue in swarmui for the extra args settingsm any quick hint?
Fdisplay advanced
Ji meant the ExtraArgs --fast you just posted under tips-info
Fit is in backends
F
Fi think speed up fp8 significantly
Jaaah thank you!
Flol some configs are pure noise
Fthis one nice but resemblance reduced
CI'm using comfyUI and the advance sample lets you choose start and end steps, and so i have two samplers in series, one takes lower model weight and passes the output latent to the other higher model weight sampler