you can search for better prompting too
you can search for better prompting too
Fbut lora may work better
Fth
Fthis config failed so bad
Fman i need 16 gpus having machine :d
Swhy you want to be bald bro?
SI tried facial scar, knife cut scar etc, it draws a bloody mess lol, but I want a cool scar like I'm some ancient warrior
Fprobably you need lora
Ftesting 
Z@Dr. Furkan Gözükara Have you tried training a LORA for an object? I'm finding when i go to generate the image using the LORA i have to describe the object in detail for it to retain most of the objects details
Fi only tried a style yet
Fit worked very well
Fdid you caption object during training?
ZYa i captioned really well, assigned trigger word, and described everything around the object i was training but didn't describe the objects details.
I wonder if i have to increase the LORA's strength when generating images...
I wonder if i have to increase the LORA's strength when generating images...
Zto be fair ive been using ai-toolkit, going to watch your kohya video, does kohya have a more up to date training method?
ZI've done it succesfully with SD, but flux is a whole other beast
Fred color blue beard almost made it fine tuning amazing
Fcaptioning your issue
Fdont caption
![Dazzastrous [4090]](https://cdn.discordapp.com/avatars/871389071686135838/c4a5f4d81d9eeb18d856295d692e2c07.webp?size=40)
Dheheh
Falso my config is very best
Fi did 104 trainings to prepare it
ZOkay gotchu!! Wont caption, and will try your config thank you! 
Forange hair white beard
Dwhat about mustache
Fi didnt set it :d
Fman deciding quality really hard between different timestep samplers
Dpink mohican
Fhaha added mustache of french i guess
F
Dvery nice
Dhipster look
F
DNow, you've become Doc Phineas from Pawn Stars
DUse that guy as inspiration for fine tuning testing lol.
Fstarting single layer training experiments
Flora
Fsingle block training speed
Freally fast
AWhen you have Quality 1 and 2 in your Flux_Fine_Tune_v2.zip file, does '1' indicate the better quality?
TWow amazing, did you use your classical 15 image dataset or the new one with different expressions
Fclassical 15 images
Fi do experiment on it
Fyes quality 1 better
AI'm about to run dreambooth for flux - I presume that we still are using regularization images, as for SDXL?
Fdoing 8x single layer training
Fno dont use reg images
Fit doesnt work
Aok
AI'll leave the reg folder location blank in kohya.
AI'm using your flux_fine_tune_v2 config.