@Furkan Gözükara SECourses have u tried using more augments apart from horizontal flip to see if the
@Furkan Gözükara SECourses have u tried using more augments apart from horizontal flip to see if they improve training?
Fwhich ones for example?
Si was thinking this, idk if makes sense using it tho, but there are plenty more https://pytorch.org/vision/stable/generated/torchvision.transforms.RandomErasing.html#torchvision.transforms.RandomErasing
Fi think trying to improve training dataset would have more dramatic effect
Slike randomly grayscale, if its 5% of the times will not affect too much
Syeah i was thinking more in terms of preventing overfitting and still improving if possible
Siam trying random image rotation -45,45 , just for the sake of experimenting
Fthanks let us know the results please
Ssure
Zi did 256 using locon on kohya, seemed to get better results
C@Furkan Gözükara SECourses I'm interested to try and run deep floyd IF locally, I have a 3090 so i should have enough RAM. I thought you had mentioned the possibility of making a video for that?
Fyep it is on my list
Fwith a good ui
CI've read that voicevox is a good opensource AI text-to-voice model but also as far as i can tell its only for japanese at the moment. Is there any equivalent for english that anyone knows about?
Mafter torch 2.o install
M
Fwell maybe kohya ss is not as optimized as sd web ui
Fbecause i know 8 gb vram people using sd web ui lora
Rwhat is the correct way to make an image with two LoRA's: 1) the subject LoRA, training for face accuracy, and 2) an outfit LoRA, training for body and outfit accuracy. In both cases, I am training for unique real object. I suppose I should treat both like you did your suit portraits using LoRA with class images, no prompt. Am I correct to suppose that I should treat both face and outfit the same way? with no captions, using class images? @Furkan Gözükara SECourses @Zono50
Fi think it should be same because you are teaching a subject as you said
Feither a person or a clothing
Mis kohya the only way to train a model well?
Fnot at all
Fthere is dreambooth extension of automatic1111
Fworks great
Mim not big into tech or anything but i need to train models well and have a 370ti
Mokay ill use dreambooth then
Mthanks
Fit has lora support
Fand works for 8 gb for sure i know that
Moh shoot ok
Mso i dnt even need to worry abt 4 gb files per model
Rhere are the results from my subject LoRA. These are ai
Fyou need extra extension
RUsing some images from a studio shoot idid
Fto generate lower size files
Fwith dreambooth extension
Fi plan to make a video but havent done yet
Fi had been waiting them to update and i think they did
Fso i can make a new video
Fmaybe tomorrow
Rbut I havent managed to get good body outfit mix yet
Fi think easier way would be
Fgenerate person with outfit
Fand inpaint face
Rthe inpaint would be using the subject LoRA only. Right. Ok
Fi got to sleep see you tomorrow hopefully
Fso many videos i need to make so little time :d
Rthank you so much! Get your well deserved sleep Dr
RSo are you simply using your Face LoRA InPainted?