Very cool, I don’t think my machine can handle kohya training but I will watch
Very cool, I don’t think my machine can handle kohya training but I will watch
Ihow do you train a lora in auto1111?
FSomething must be wrong. You are using realistic vision v2? With man word?
FVery weird you are using text 2 img?
FI don't use captions for faces. I find that they degrade teaching a person
FThe logic is different. You can watch my this video to understand token logic https://youtu.be/Bdl-jWR3Ukc
YouTubeSECourses
Our Discord : https://discord.gg/HbqgGaZVmr. The most advanced tutorial of Stable Diffusion Dreambooth Training. If I have been of assistance to you and you would like to show your support for my work, please consider becoming a patron on 
https://www.patreon.com/SECourses
Playlist of Stable Diffusion Tutorials, Automatic1111 and Google Colab ...
https://www.patreon.com/SECoursesPlaylist of Stable Diffusion Tutorials, Automatic1111 and Google Colab ...
FThank you so much for the amazing comment
FI haven't compared yet but planning to do
FYou can continue. It is just slower. Only important thing is xformers version. Xformers 0.0.16 don't work
Ii have 0.0.18
FYou are welcome. It supports cpu training too. How much vram you had?
FOK it works 
FHopefully I will make a video waiting them to merge dev into main
D@Furkan Gözükara SECourses Possibly another issue for me, I am not running "Darwin" debian but a debian derivative, POP_OS by system76, so when I run the "upgrade.sh" script it errors out. I'm going to look for another solution to train on a GNU/Linux system. Maybe one with more detailed instructions to make the system work properly. Thank you!
FYes hopefully I will make a tutorial for Runpod. It uses Linux so you should be able to follow instructions
Iwhy is torch 2 slower on kohya ss?
FYe that is interesting. Maybe because I tested with Fp16 and not bf16
FCan you also test?
Ii selected no there
Ii have 24GB VRAM
FYes makes sense
FMaybe that is reason
Ialmost ready to train, working through your tutorial, thank you
Ihow are you running stable diffusion and kohya ss at the same time, it told me i could only run one at a time
Invm got it
IHey guys, i need to use SD+ControlNet+ Temporal Kit, do you have any colab i can use please? i used colab only for dreambooth
FChange port
FUnfortunately don't have colab for that :/
IThanks anyways, i'm doing some test for making a decent video but i need more vram
Imy 8gb are decent but with more i can have more stability
Ilol now to gen 600 images of "man"
IIthank gods for 3090
DI'm going to try InvokeAI and see how that works on my machine. So far I like it's installer. We'll see how it goes...
JI don't find the same at all, without captions, it's very difficult to try and align the training, seems to wonder all over the place
CHey @Furkan Gözükara SECourses, are you open to consulting inquiries?
CRunning a commercial stable diffusion inference pipeline and we need our DB training audited.
JI used unipc and cranked out a 100 batches of 5 at like 10 steps
JIt worked pretty well
JTook about 15 mins
Ii just did default and it genned 600 in about 5 mins
I20 steps, euler a 512x512, not bad
Iim only to the settings part now lol
Iif i train a single lora before bed ill be happy
DWhy not just download the images of Man that Dr. G. made available for download as a Zip to save time, or are you testing your system?
Iam i doing something wrong? its saying its going to take 1.5 hrs, i thought the loras only took like 10-15 minutes?
Iit said it was behind a patron wall
F8 gb good for lora
FYou are teaching a person? How many images you are training
FYes I am
