Dbut you are using adamw8bit
Dit consumes more vram
DI use adafactor
Dwhat graphic card do you have?
MAnyone use Prodigy for Flux traning?
Dnot me, it too memory hungry
Sare there any differences i should be aware of when making a gguf of a flux.1-dev fp32 safetensor model, compared to the way i would do it for an llm? just looking to make a q8 for people with less vram.
Jusing only original ones from Black forrest labs (Flux1-dev),
there are too many versions out for me to keep track of and whether they are all
compatible with my LoRAs or not.
there are too many versions out for me to keep track of and whether they are all
compatible with my LoRAs or not.
TYeah they are original, you can read the repo page to have the info
N@Dr. Furkan Gözükara 
Congrats on the review, will peruse and need to put likes everywhere
Congrats on the review, will peruse and need to put likes everywhere
Fprobably you loaded into dreambooth tab at least once and corrupted config
Fyes this is a bug
Fyou need to use command
Flike cd checkpoints
Fi didnt try lion
Fyou have to use cfg 1
Fflux has different guidance
FWell 102 is decent number so even if you go 100 epoch it makes 10000 steps 
FI need to test it and see if works
FRegularazation I tested it. It will make your face mixed
F1 repeat 200 epochs
FI trained turbo schenell model
FResults are bad as expected. It trained though
FCurrently training on 4x gpu to see its quality - practical batch size being 4
DI saw it in your video, you tested again without class with the same results?
Fi trained only ohwx but didnt use reg that time
Fcan you explain me exact setup you meaning
Idoes your tutorial give better results than training lora on site like replicate?
DI was thinking to train only ohwx for your pictures with a high TE learning rate to link the token ohwx to you and the regulation to the token man
Dthis may be can reduce mixing
Fit gave better than civitai
Fi didnt test replicate
Foh i see probably wont work. because flux itself still tags me as man
Fits inner text encoder alike attention
F4x gpu results are great as well
Falmost linear speed increase
Fgonna record runpod and massed compute tutorial hopefully today
Dyes but it worked for the base model training, celebrities are women and men and no end up mixed
Dsomething is missing, may be something is wrong with the clip_l te training, or it needs higher LR
CI still get the same error even after uploading all files to the folder
Dthe other thing that can be is that is a distilled model and something is missed during the distillation and cant link new tokens to a new unique concept like a especific person without overwriting the entire class
Ftrue i thinkn key might be fine tuning
Fyou want to install swarmui?
Fwhat you want to do?
Fgood point
Fi added 2 new research results
FPatreon
Get more from SECourses: Tutorials, Guides, Resources, Training, FLUX, MidJourney, Voice Clone, TTS, ChatGPT, GPT, LLM, Scripts by Furkan Gözü on Patreon
CPatreon
Get more from SECourses: Tutorials, Guides, Resources, Training, FLUX, MidJourney, Voice Clone, TTS, ChatGPT, GPT, LLM, Scripts by Furkan Gözü on Patreon
