hello. are you using latest automatic1111? this was fixed
hello. are you using latest automatic1111? this was fixed
Fyou should do pip uninstall torch torchvision torchaudio then reinstall
Fjust scribble and prompt may work
AThey resolved it by updating to the latest version 
ASo far in my limited testing I am seeing better results using
Adan Dadaptation than Lion optimizer, but the dev branch of Dreambooth requires way too much VRAM.
Finteresting i wonder what caused the spike
AApparently not tested with Linux 
Arrow — Today at 10:42 PM
I haven't ran this with linux, so I don't really know. You'd probably have to search the web for the issue and tinker around. A quick thing you could try is disabling torch dynamo at the top of dreambooth/train_dreambooth.py:inner_loop
https://pytorch.org/docs/stable/_dynamo.html#torch._dynamo.disable
Arrow — Today at 10:42 PM
I haven't ran this with linux, so I don't really know. You'd probably have to search the web for the issue and tinker around. A quick thing you could try is disabling torch dynamo at the top of dreambooth/train_dreambooth.py:inner_loop
https://pytorch.org/docs/stable/_dynamo.html#torch._dynamo.disable
ZI've gotten it to create the images and upscale, but at the end ffmpeg does it's split, i get an error
unrecognized option 'crf' .
unrecognized option 'crf' .
message.txt2.72KB
Fi use davinci resolve
Fto reconstruct
Ffree edition
Zi'll see if i can install that on manjaro
Zso is it safe to say that the built in reconstruction using the gpu option is broken at this point?
Fgpu working for windows
Fon colab working as well
Zyeah idk why it won't reconstruct it on mine, followed all the reports that I can see, unfortunately they closed the discussion group
Zyou know me, i'll figure it out, will just take some time
Fyep you are really good
F@Dr. Furkan Gözükara did you change the tutorial for face swap?
Fwhere is the -f and -o parameters?
For they were never there?
Fthat is for google colab
For if you run without gui
Fnow works with gpu too
Foh right!
Fi got it from colab not the github tutorial
Fi can't get cudnn to work on runpod
Fit is much faster with gpu?
FGitHub
Regardless if you guys have pulled the model for any which reason, it's probably best to verify the hashes below so that people don't download anything malicious? Below is the version I dis...
Flet me do test
Fgpu : Processing time: 21.62 seconds
F-- max core was 1 : Processing time: 138.24 seconds
Fouch xD
Fand without max cores ?
Fcan you try please? on runpod i don't limit cores
Fye this guys are not pro
Fi can try max core but afraid of blue screen :d let me test
Fi use max cores and no blue screen, it just makes my pc slow XD
Fmax core really problematic i had to terminate
Fwill try with 8
ZWhat version of ffmeg you guys using?
Fffmpeg version 6.0-full_build-www.gyan.dev Copyright (c) 2000-2023 the FFmpeg developers
Ftried to termiante process and got blue screen
Fmy commit id i think not supporting multi threading actually :d
Zoh god, i'm using pips version 1.4
Zsystem version is 4.4
Fi will make a wild training test tonight
F1024 pixel over 100 images to teach a person
Fdoing a very special training first time
Zdoing kohya?
