Is this the result of the 256 image data set + 7 batch size?
Is this the result of the 256 image data set + 7 batch size?
F
Fit is around 1460 steps and 1 step is around 30 seconds
Fon single A6000
Fso cost is around 3 4 usd
O$34? or 3 to 4
F3 , 4
F31 cent per hour
Fyou can calculate easy
F1450 * 30 second = 12 hours
Osounds about right
Fi am analyzing around 1000 stylize prompts to note good ones
Fby the way 256 images is overkill 
Ois there a trick to getting better faces at a distance? i.e. not just portraits?
F
Finpainting
Osigh, lol
F
F
F
Fnice one
Fwill put this into realistic prompts
Fnice it follows some stylies good
F
F
![Dazzastrous [4090]](https://cdn.discordapp.com/avatars/871389071686135838/c4a5f4d81d9eeb18d856295d692e2c07.webp?size=40)
Dwhat about clay model
Fnice
![Dazzastrous [4090]](https://cdn.discordapp.com/avatars/871389071686135838/c4a5f4d81d9eeb18d856295d692e2c07.webp?size=16)
Flet me test
KIs this from your latest fine tuning config from Patreon ?
Fyep batch size 7 fine tuning
KDo we have to enable prefer sysmem fallback?
Fi have it enabled
Fdidnt change
Fbut you shouldnt use shared vram
Fduring training
Fhaha this one nice
F
F
Fclay is hard
Fneed to find a betterr prompt
F
F
Fpretty good
Dfurkan action figure 
Fye but really hard we really need to reduce bleeding
we really need to reduce bleedingF
Xamazing!
C@Dr. Furkan Gözükara dont you have family or friends to train a lora on so its not just all you?
Fwell of course i have but i couldnt collect 256 images of my family. also purpose is finding workflow and configs
