Training dataset, different shirts, different backgrounds
Training dataset, different shirts, different backgrounds
Fhow many per epoch do you save
Falso have you tried different prompts?
GI have tried MANY MANY MANY propts and I think it was 10 per epoch
Fin your example | didnt work
GCorrect. It did not work but it was the last thing I tried before deleting my pod
Fi see
Fif you are our patreon supporter i can setup training for you via connecting your computer
GThat sounds like a good idea. I might look into that.
Scan anyone please help me with loha training T_T
Fwhat help you need
SIdk which parameters i should be using
Shere is what i have rn
SLoHaBasicSettings.json2.12KB
Fye cant help with that. it requires extensive experiment and i didnt make yet
Fhow much vram you have
S16
Fvram or ram
Svram
Fwhat is your gu
Fgpu
S3080ti
Fit has 12 gb vram
Fat least
Fi suggest you to use dreambooth
Fextension of web ui
Fworks great
Sbut lora faster and also weights less no?
Fyou can see my extensive test here
FYouTubeSECourses - Software Engineering Courses
RunPod: https://bit.ly/RunPodIO. Discord : https://bit.ly/SECoursesDiscord. Web UI DreamBooth got epic update and we tested all new features to find the best settings (10). If I have been of assistance to you and you would like to show your support for my work, please consider becoming a patron on 
https://www.patreon.com/SECourses
Playlist of...
https://www.patreon.com/SECoursesPlaylist of...
Fcorrect
Fbut much more variables to configure in lora
Q@Furkan Gözükara SECourses do you think dreambooth is better then hypernetwork training in web ui?
Sdreambooth will give me a checkpoint right?
FI think dreambooth extension has matured greatly
FYes 4gb safe tensors file
Shmm yeah but then i wont be able to create my characters in different models
FWith lora you still won't be very much
FEach training only works best on trained model
Shmmm but I've tried different loras on multiple models and they all give a great result and now with the new update to lora which is loha
Sthe results r amazing
Sand even half the size
Fthey must have common root model probably
Syeah ofc the sd1.5
Fso you tried default settings? i plan a tutorial but still couldnt find time yet
Soh yeah i can do normal lora training and the results r good but the loha one is new and I cant find many tutorials since it just came out recently but the ppl who have made some r giving great results even with default settings
SGitHub
Lora beYond Conventional methods, Other Rank adaptation Implementations for Stable diffusion. - GitHub - KohakuBlueleaf/LyCORIS: Lora beYond Conventional methods, Other Rank adaptation Implementati...
Fi will ask developers
Fok extended lora in dreambooth extension of web ui already supports these
II tried to replicate your last settings today, and for the first time I got complete noisy images.