yes that's very hard and how do you comapre if the previous models were better? are you uisng same s
yes that's very hard and how do you comapre if the previous models were better? are you uisng same seed, same prompt, etc?
Fi generate image of previous model as well in same grid
�okay. I trained a dreambooth yesterday on that outfit I asked you about but I think the learning rate you set is very low should I increase that to 1e-4 ?
Fyou can increase if it is undertrained
P@Dr. Furkan Gözükara I have few questions: 1)The hands especially the hands and fingers isn't good on my flux+dreambooth trained model ,how to fix that ?2)as we had trained the model for 1024x1024, what if I want to generate image of high resolutions like full HD or 4k 1920x1080, 3840x2160?
Fupscale images it really fixes a lot
F2x latent upscale
FPAnd what about increasing resolution will it do that too along with fixing the hands?
Ayes also for the faces too
AI've tested it
Abut it will be slower
Fwell i think it depends on dataset you can test
Fnormally i have no issues with hands usually
For i generate more images and get perfect hand :d
�@Dr. Furkan Gözükara In the 48gb dreambooth config do you have text encoder training on?
FYes. I have installed everything like mentioned in the windows prerequisite video. also I tested with LivePortrait after installing main things. Its working fine. Issue is I'm unable to install Kohya (both Flux and Dreambooth) using 1 click installer
�@Dr. Furkan Gözükara A good read about loss values - https://github.com/spacepxl/demystifying-sd-finetuning
GitHub
Contribute to spacepxl/demystifying-sd-finetuning development by creating an account on GitHub.
Fwill check sorry for delay
Fye i read it
Fyou are using v17 right?=
FPatreon
Get more from SECourses: FLUX, Tutorials, Guides, Resources, Training, Scripts on Patreon
FYes. Using v17
Fgui started for me
Fcan you tell me folder path you installed?
Jlets goooo
Fis gui not starting or you get error at training
FI'm getting error while installing
Jdo you recommend just using this instead of SUPIR? I'm guessing supir might be slightly better but not generally worth it?
Fcan you make a fresh install and send me install logs
Fsupir is best to upscale existing images
Fbut if you fine tune flux latent upscale works even better to add more details and fix errors
Jhuh okay gotta try that, thx!
Fsure. where can i find the install logs?
FTried new install. Now this command prompt closes abruptly after that line "creating venv".
Finstall logs means the output on cmd which you can copy
Fwell restart pc and try again
AGuys if I want to remove false letters and labels from a shirt for example , how do I prompt with flux fill dev so that it removes the letters and keeps the texture/ color of the shirt?
Dbrother is there any RVC Retrieval-based Voice Conversion tutorial?
FResolved it by running this command git config --global --add safe.directory D:/AI/Kohya_Dreambooth/kohya_ss
Fsadly not yet
Fah it was permission related
P@Dr. Furkan Gözükara when I am using the upscaler you told to use the generated images are looking not realistic
Pthey are get bad textures like dolls
Pthe hands and other things did improve but the skin isnt' looking fully real
Fyou probably need bigger training dataset
Fit improves realism
Falso are you using my said upscaler?
Pyes
Fim an admin user. I dont know why it was making an issue
M@Dr. Furkan Gözükara I've watched the tutorial on Flux Tools in SwarmUI, and I have a question.I see that you generate the image directly with the model, meaning you load canny.dev or redux.dev instead of the original Flux model. These are tools that, in theory, are compatible with the original Flux models. So, could I use flux-dev together with flux.redux to get the most out of it? Or is there something I'm not understanding?
