I had the best result at 400 - sadly these are very old family photos, so not enough available photo
I had the best result at 400 - sadly these are very old family photos, so not enough available photos of my parents 
AI think after 400 or 500 it starts to overtrain.
Ri think there is a complier error
Fvery likely
Fcuda 12.4 + accurate c++ tools you need
Fye it depends on each dataset
F400 not too much
Ras far as i can tell i have both
Ffor 10 images
Fbut inaccurate versions very likely
Rhow do i check
FYouTubeSECourses
If you are interested in using AI, generative AI applications, open source applications in your computer, than this is the most fundamental and important tutorial that you need. In this tutorial I show and explain how to properly install appropriate Python versions accurately, how to switch between different Python versions, how to install diffe...
Fi am still using this video config
Ri did that. not sure what i missed
Ris there a way to check what i missed installing
Fsadly not
Fthese compiler stuff so so hard to solve
Fsometimes i am spending 10 hours to make scripts work
Ri see. it just c++ stuff i have to fix right,not everything
Rwhy can't this be a docker image
Rlol
Ri used cuddnn 9.5.1 is that fine
Rit the lasstest
Ri reinstalled everything
RI honestly at a loss, so i am giving up for now. I can't figuer out what i doing worng
Ri just installed like every build tool in visual studio
Fi feel you. you need to invest days maybe months to learn all these :9
Fi am giving private consultation if you are interested in
๏ฟฝI just made a lora from a fine-tuned flux dev checkpoint, it's size became 11.3 gb. Can that really be the size of a lora? 
F640 rank float can be :d
Rsure
๏ฟฝHaha yes I went with High rank, so I get higher quality, but also slower rendering and loading.
Ri feel like i am downloading the libarly of congress with this models
Rand the download speed is so slow
AWell, a 640 rank ought to be enough for anybody 
/BillGates
/BillGates๏ฟฝFor a face only lora, which I will use in the <segment:yoloface...> part, 15-20 pics should do it yes?
๏ฟฝdifferent angles, expressions and lighting
RI am unsure when one should use dreambooth vs lora
Fwhat is your purpose?
RI want to generate images that make up elements of differnt people but not recoreate a spefic person
Fso you want to train a style
Ffine tuning better
Ris it a style, like what defines a style
Fyep do fine tune
Fdreambooth
Rcan you clarafy the types of models. like styles,subjects.
Rwould i use dream booth for recreating myself as well
Rcan you continue training from a model you trained already. if so how, because i keep running into issues with that.
Fyou need lora only when you want to publish on civitai so that people can use on different base models - otherwise lora is no, it is inferior from all aspects. also you would want lora if you are service so you want to host small files
