Should I lower the steps or the number of images?
Should I lower the steps or the number of images?
Fyou should fix setup
Fwhy you have that many steps you know?
FI have no idea
FMaybe because I clicked on prepare training data?
FThis is what I have
Kohya_ss_GUI_v24.2.0.pdf1.86MB
Fyou have 40 repeats in config
Fshow config folder names
Nseems like discussion here is usually around only 10-20 training images, and that by 1000-2000 steps the lora produces a great likeness (assuming you're going for a character lora). since you have 3x-6x as many training images, you'll have 3x-6x as many steps... steps is images * repeats * epochs. the doc's configs always have repeats at 1, so it's just images * epochs. might be worth additional testing to loras made by 10 images at 200 epochs (2000 steps) vs 20 images at 100 epochs (also 2000 steps). if we increase images, do we get a "good" lora faster? or if we increase images, do we need more steps to better "merge" all of them? i am just a hobbyist and now i'm rambling lol - would be interested in the Dr's feedback.
Fi hate mountain house electricity went off :/
Fi will test more images better or not with flux but still optimizing hyper params
Freg images definitely not helping
Fi am going to sleep probably electricty wont come until morning :/ 2 am here
Fhaha electricty came faster 
GWhere you at bro? South America?
FTurkey
Fin mountain house
GI thought you were Hispanic
Fnope
FNgood news about your electricity - in that case i have more questions lol
in your SDXL fine tunes, you train off the RealVis model instead of the SDXL base
there are several fine tuned flux models on civitai with thousands of downloads already...
should we be able to use those as a base like you did with SDXL? should flux loras work across models?
in your SDXL fine tunes, you train off the RealVis model instead of the SDXL base
there are several fine tuned flux models on civitai with thousands of downloads already...
should we be able to use those as a base like you did with SDXL? should flux loras work across models?
Gbro how many times has the power went out during trainings? haha that would suck
GI hope you have good UPS to protect your GPU's
Ni'm thinking when he's ripping off 8 trainings at once, that's in the cloud lol
GI run step 1, install Kohya, then Install CUDNN, then step 2 fix for flux
Gis this fine?
G
G
Gstep 2 finishes
Ntry it and see?
Fi am training on massed compute
F
Fprobably those models are very weak
Goh no power outages haha
Fah let me update file
N@Dr. Furkan Gözükara are your sample prompts on patreon anywhere? do you use a wild card extension? would love to try my loras against some of your same prompts.
GI do install CUDNN right? after Kohya? And ignore steps 3-6 right?
FPatreon
Get more from SECourses: Tutorials, Guides, Resources, Training, FLUX, MidJourney, Voice Clone, TTS, ChatGPT, GPT, LLM, Scripts by Furkan Gözü on Patreon
Fi, just added sample prompts
Fif you use grid you need to merge them with ||
For you can make them wildcard in swarmui
Fdownload v14
Fjust 1 and then 7
Falso you can set accelerate
Fshown in new video
FYouTubeSECourses
Many people having issues due to inaccurate Accelerate setup. This is how to setup to make it work properly
1-click installers and best configs for Kohya FLUX training :
A Big complete tutorial very soon hopefully : https://www.patreon.com/posts/kohya-flux-lora-110293257
1-click installers and best configs for Kohya FLUX training :
A Big complete tutorial very soon hopefully : https://www.patreon.com/posts/kohya-flux-lora-110293257
GOk will do 1 then 5 then 7 then Flux fix 2 batch
Gthen Start Kohya right
Fsounding good
GI thought accelerate was for low vram cards, it will work well with a6000 too?
Fit is for all
