ram usage shouldnt be much unless your vram is low
ram usage shouldnt be much unless your vram is low
Fbecause when vram is low cpu off loading is made
Fand it uses a lot of ram
Fwell EDM is more loyal but you can use both 
Factually we are working on adding a lot of new stuff too
Fhi are you using gradio live share?
Fvery likely your runpod pod is broken
Fi would prefer massed compute over runpod anytime
AApreciate your prompt response, i am using direct http link from pod
AI will give Massed compute a try.
Fah yes that is why failing
Frunpod proxy
Fwow this is really great for flux
F
BAnyone here got JoyCaption working in ComfyUI?
OJust to clarify: no multi-gpu, but an A6000 can skip tiling because it has 48gb vram?
OIs this the result of the 256 image data set + 7 batch size?
Fyes
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
