let me update the notebook to v23
let me update the notebook to v23
GYeah I'm sure:
Glegend 
Ftrying to fix
Fthere is an annoying gradio error 
Fye not working atm i opened 2 issues
Fi got to sleep
Gno problem at all, appreciate you taking a look at it
GI managed to hack the notebook and get it working by avoiding gradio altogether (just manually creating the input folder structure) so no longer a pressing issue
IHi! Is there already a lesson on recognizing styles? In the lesson on learning Lora ( https://youtu.be/nySGu12Y05k?si=zQscauo48UK54N1F&t=1519 ) you said you'd do it. Maybe I didn't search well and didn't find it, or maybe there wasn't a lesson yet. Tell me, please, is there a separate lesson on image style recognition? This refers to Lora, which copy the style of different images, rather than the style of one person or object
Fye kohya itself should work fairly fine
F
SGuys i'm new here, just wanted to know what system configuration is Required to run Kohya ss Completely Locally
PhI @Dr. Furkan Gözükara my client is ready to accpet current results with 220 images flux dev training ,i sent them the grid ,I have 10 checkpoints ,the client says they want the whole 10 images that is 1 image per checkpoint as the output and they will select the best ,so how to host the whole 10 models on the cloud and it should run 24*7 without my need and the output should be the whole 10 images for each prompt ,whats the method to do this and host it in affordable price.The total size of the 10 models in 240 gb
Scheck terabox
SOr you can purchases One drive Monthly plan
Fhey welcome
F64gb ram and 6gb or above gpu
P@Dr. Furkan Gözükara how to make one prompt generate multiple images in one go like chatgpt ,so that I can chose the best one ?
Fyou can increase number of images generation
Pwhere is that option in swarmui?
F
Ffrom first 5 trainings
Fearly results of SD 3.5 research
Fclip g and clip l training supported but they are seperate files. i talked with kohya. i hope he implements
Ano way near flux at all
Ait's decent 7/10
Fye nothing sort of flux but hopefully this will get way better
Fi am just starting to research
Ahow's the skyreels tutorial going ?
Fi plan to record today if i can make it so many stuff
so many stuff
Ndr just wanted to point out something happening very regularly for a month now on Massed Compute
Nwhen using your SECourses launch
N1 of 4 instances sometimes all of them just don't work
Ntheir pc is always booting and never ready to launch
Nbut with other instances it open as regular
NI guess there is some scripts inside that sometimes work and other times don't that cause that problem
Njust wanted to point it out
N@Dr. Furkan Gözükara
Ni think it is the updaters of the programmes that you got on desktop, cause sometimes if I open the instance before is "ready" i see the errors on them that causing the problem
and sometimes if I just restart the computer on ThinkLink client, only then it works
and sometimes if I just restart the computer on ThinkLink client, only then it works
Fsome important information lets ping @NicB @computersheep
Flol looks like adafactor is still best i did huge trainings to find better optimizer but it still cant pass adafactor
i did huge trainings to find better optimizer but it still cant pass adafactorFi will try prodigy
Fnice i made ProdigyPlusScheduleFree made work testing
F3 FLUX Dev 3 SD 3.5 Large research trainings running
Fman sd 3.5 large is weak
Fthis is one of the best images so far
Twhy do you focus research outdated model?
Tah, 3.5 large is a new model, lol
1
1API I2V was impressive
