i am adding multi gpu to joycaption app
i am adding multi gpu to joycaption app
Feven flux authors didnt train it
Fso unnecessary
KThis one here
KJoycaption... Gotta try that one
KIs it fast?
Fdecent under 10 seconds
Fi will caption dataset on 8 gpu 
Ftrying to code multi gpu
KHe he... I only have one 
Fit doesnt have example gradio there
HGood evening! When I create a pod through RunPod using the Stable Diffusion template with web UI, But when connecting the pod, the option to access Jupyter Lab is not ready.
Frefresh and try again
Fit make take time
Fif never becomes delete pod and make a new one
HThank you!
HIs the Expose HTTP Port correct?: 8888,3001
Ddid you post some of your prompts anyware to test myself?
Fyep
Fprompts included in patreon zip file let me update to latest prompts
Fi will also make a new post
Fmore tidy
FPatreon
Get more from SECourses: Tutorials, Guides, Resources, Training, FLUX, MidJourney, Voice Clone, TTS, ChatGPT, GPT, LLM, Scripts by Furkan Gözü on Patreon
Dthanks!!!!!
Sgod bless you Dr Furkan
Dwhat is the ideal max new tokens value to prevent hallucinating? Also I installed V9 in windows, Is there any way I can simply upgrade it to V10 without re-installing? (Joycaption)
Dwhat is the prompt input instruction being used in Joycaption by default? Can I customize it?
Fjust overwrite app.py file
F300 tokens i think good
Fno instructions it is fine tuned
Fi am working on multi gpu support
DWhen is Supir new local version launching (V2), any dates?
Fnope not yet
Fman multi gpu app is hard
Ftesting multi gpu joy caption
Fgonna caption 20k images :d
WQuestion. Is it possible to train a Lora on 2-3 images of a person if that’s all I have? Also how many should be used to train a style Lora?
Fas many as possible for style
Ffor sdxl 200 was decent
F2-3 images may work but i didnt test
Fusually you need to have different poses distances and expressions if you need
FI'm trying to get SwarmUI working and it's not being very cooperative
Fmake cfg 1
Ffor flux
Fi made joycaption multi threaded
Fit is slower than 1 thread but 8 gpu works great
FTy, now the result looks correct
F
FTesting another LoRA
