i think so, very well explained
i think so, very well explained
Cis this a good place to share results? or is there like a show and tell room?
Ci'll see..lol it might be a while until i have anything great
Yfaster then you think !!
Fxformers versions updated for torch 1.13 : https://github.com/FurkanGozukara/Stable-Diffusion/blob/main/Torch-1.13-and-xFormers-19.md
GitHub
Best Stable Diffusion and AI Tutorials, Guides, News, Tips and Tricks - Stable-Diffusion/Torch-1.13-and-xFormers-19.md at main · FurkanGozukara/Stable-Diffusion
Fespecially important for runpod
Mhere's a question: I'm trying to create a lora of a type of humanoid (Disney gargoyles: https://en.wikipedia.org/wiki/Gargoyles_(TV_series)). I followed this tutorial to make the lora and I used a token like "d4m1 tiefling" and my reg images were based on the class "tiefling." Tiefling are fantasy race similar in appearance to gargoyles so I figured that might be a close match. I can't say tho the lora is really drawing from my training set. Does it make more sense to do a class like "man" and then one with "woman" instead?
M(or is this more a task for fine-tuning?)
HHi. Can I run stable-diffusion with dreambooth on python 3.11 with torch 2.0.1+cu118 and without installing xformers ? What do I need xformers for? (win10)
ADepends on how much VRAM your GPU has.. low VRAM requires xformers but better to disable it if you can because it has a negative effect on training quality.
H8G on the graphics card plus 32G shared mem
AThat's not enough VRAM for Dreambooth unfortunately. You should check out RunPod.
Hhmm, okay thank you. Playing around with this since yesterday. I think I'll go shout into my pillow for a moment and then watch Netflix.
Estill a little new here. I am trying to wrap my head around why epochs are halved if using classification images in kohya lora training (shown here https://youtu.be/TpuDOsuKIBo?t=1456) should i just not worry the "why" about it? lol also, im doing a training right now, training myself as the subject. 22 images, did 40 repeats, 10 epochs, so i thought that should equal out to 8,800 steps, but its actually showing a total of 17,600 (double) is that because i have class images?
YouTubeSECourses
#Kohya SS web GUI DreamBooth #LoRA training full tutorial. You don't need technical knowledge to follow this tutorial. In this tutorial I have explained how to generate professional photo studio quality portrait / self images for free with Stable Diffusion training.
Our Discord server
https://bit.ly/SECoursesDiscord
If I have been of assis...
Our Discord server
https://bit.ly/SECoursesDiscord
If I have been of assis...
Fye you may need to try different classes
Fno python 3.11. use python 3.10
Fwith 8 gb you can do lora training
Fi think they might have fixed it.
Fdouble steps are correct
Fthat is a fix
EI want to do no more than 6,000 ish steps for Lora training a face, right? So does that mean 12,000, or 6,000 
?
?Fyou should aim epoch
Fnot steps
FI got kohya scritps working today. very successful with them. I used fp16 instead of bf16
Yi would be glad to see a tutorial about the training on kohya ss guy in case automatic 1111 broke again
Fye bf16 not available on all cards
Fi already have tutorial :d
FYouTubeSECourses
#Kohya SS web GUI DreamBooth #LoRA training full tutorial. You don't need technical knowledge to follow this tutorial. In this tutorial I have explained how to generate professional photo studio quality portrait / self images for free with Stable Diffusion training.
Our Discord server
https://bit.ly/SECoursesDiscord
If I have been of assis...
Our Discord server
https://bit.ly/SECoursesDiscord
If I have been of assis...
Ylol thx i will see it
Fi have planned so many tutorials but so many things taking so much time including family :d
Frtx 4080 it should work
Fyes i used the one linked on SECourses and a few others here and there. kohya is very widely supported . most tutorials i've noticed are for it, rather than the webui
Fyes
Fi also managed to get pytorch 2 working so i trained a style using lion. got some nice results but i might've set the lr way to low
Fi use 1e-7 for lion
Ftorch 2 works on windows
Fbut on unix i couldnt yet
Ysimple question : i've trained a dreambooth that work fine as long i don't put " oil painting in the prompt : in this case the face is destroyed
EI figured epoch is more important, but then how would I decid how many repeats? :/ ugh I feel that dream booth extension is simpler, but it always gives us trouble.
Fthat means either overtrained or undertrained
Fye it depends on used script
Fnormally 1 epoch means 1 time all training items are trained
Yi have several steps ckpt of this model i will try them all
Fdreambooth learning rate 0.00001 vs 0.000002 anyone?
Fi prefer 1e-6 for adam, 1e-7 for lion
Yi use 0.000002
Fbut you tried 0.00001 ?
Ynot so far
Mif we’re trying to train concepts (again I’m doing fantasy races), does it make sense to include a lot more training images?
Fquality over quantity i lean towards. add more but only if they meet a bar
Fthe concept in a myriad of ways too . not more images that are similar to whats there already. unless they're high quality haha
