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Twhy do you focus research outdated model?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
Fthere is a client requested
Fyes but code missing :d
Foh 9 min ago published
Fworking on installer
Fdmn still no torchvision for rtx 5000 series
1I didn't expect the 14b to be released too
Fthey released both amazing
Fmaking the gradio app installer
1
1I'd be interested in helping to test
Ftrying to make installer
Fdmn it uses flash attention forcedly
Fi installed weird testing again
Fok downloading models fingers crossed
Fgenerating video
F
1looks promising
Fafter prompt enhance it doesnt deload model lol
F
Fvae decoding taking forever
Fexcellent
F1ohh damn, that's only the 1.3B model
Fyes
Fdownloading 14b model for image to video
Fbut i dont know if will fit vram