I had to run one of the .bat files to avoid running out of memory on a 4090 - the file for installin
I had to run one of the .bat files to avoid running out of memory on a 4090 - the file for installing Torch 2.5.
Fyes it is amazing improvement
AOne odd thing - I was using ComfyUI to generate some 2048 x 768 banners, and I started to get some fine vertical lines appearing in the imagery, which disappears if I reduce the resolution.
Fthis is known issue
Fi get same
Fat higher res
AAlso, all of your gridded comparison graphics - outputs for various parameter ranges - are so handy to refer to.
XDoes Model Downloader inside of swarmui not work all the time?
Fsomeone were also reporting similar stuff
XYea it seems to not be working for me
Dis it possible to define a custom mask in swarm instead if using auto segment?
Fit supports many classes
For yolo models
Flike you can do
F<segment:car>
Dis it possible to invert the segment class?
Fyou should ask that
FDthanks
Xi asked them, i ended up just manually downloading it
Xwhen using multiple loras ex.https://civitai.com/models/647663/porsche-911-gts-2024-flux?modelVersionId=724571 & my ohwx man how could I prompt for the car?
LoRa of the Porsche 911 GTS 2024. Generation tipps: You can generate directly 1920x1080 without problems. It's a bit overtrained but retains more d...
Xlike which trigger word should i use for the car lora 
Fi never tried :d
FPatreon
Get more from SECourses: Tutorials, Guides, Resources, Training, FLUX, MidJourney, Voice Clone, TTS, ChatGPT, GPT, LLM, Scripts by Furkan Gözü on Patreon
Fthis is how you install stuff on massed compute
X
Xused 2 loras "ohwx man" & "Porsche 911 Turbo S" i think it looks preety good. It has no upscale also
Fearly fine tuning results,
F
F
F
Fthis is with 15 images dataset
Fmaking cartoon still broken 
F
Fenviroment improved
F
Xthe image enhancements always looks great on your generations'
Fthese are raw only face inpainted
F
Freally good armor quality
F150 epoch
Fpretty fast training also
Fit has 0 overfitting from my full overfit dataset :d
HAnyone have any ideas on how I can take a realistic face i made on an SD 1.5 model and make that same realistic character on flux without training? Reactor extension leads to subpar results. I know it will look a little bit different but I want to maximize the resemblance
Finstantid your best shot
Fwe have very good app for it
FPatreon
Get more from SECourses: Tutorials, Guides, Resources, Training, FLUX, MidJourney, Voice Clone, TTS, ChatGPT, GPT, LLM, Scripts by Furkan Gözü on Patreon
Xwow i need to do more training might be better for me to get on mased compute