i plan to compare all optimizers in kohya lora
i plan to compare all optimizers in kohya lora
AI removed from prompting.. I still use the class images.. but removing girl from training solved all the problems that I've had for months lol... I still don't know why it was such a specific problem with my one daughter.. probably because she var very hard for the model to determin her age.. my younger daughter who is 5 it had no problem with..
Fi thinking that this crappy regularization images that i'm using are deteriorating the results
Fthe woman class is re-trained with monsters
Fcould be nice find
Fcould be
Fi'm gonna try filewords + controlnet, someone told me that he's having a great result generating class images with controlnet depth map
Fso class images replicate the training images the most possible, but without the person face bias
Aif you buy multiple GPU's can use stack the VRAM to make training one model go faster?
like can you buy multiple 3090's in one computer and use them all on one A1111 image/model training?
like can you buy multiple 3090's in one computer and use them all on one A1111 image/model training?
Thow is amd looking for training right now? i want to buy a 7900xtx 
DHey guys what is the least amount of regularization images i need for artstyle and man?
DLike I am generating my own to train on top of a specific model so I can't wait for 300
DDaaamn i9-1300k did you also upgrade your gpu?
DMan I can relate to this so much it's actually spinning my mind now
KThat's awesome I can't wait to see your new video. And btw, in Kohya Dreambooth, what is the difference between "30 repeats x 10 epochs" and "10 repeats x 30 epochs"? Which option would produce better quality, or are they the same?
AI assume that its better for all of the images in the training data set to be in colour than to have 1 or 2 black and white ones?
Aany fix for this ?
AAre you creating your models from Hugginface, by checking this?
AIf not, then you are trying to trying to train a model before actually creating it. I had that happen to me yesterday.
ANo I using the pre-trained model, shown in this tutorial https://www.youtube.com/watch?v=kIyqAdd_i10
YouTubeSECourses
Our Discord : https://discord.gg/HbqgGaZVmr. This is the video where you will learn how to use Google Colab for Stable Diffusion. 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...
https://www.patreon.com/SECoursesPlaylist of Stable Diffusion Tutorials, #Automatic1111...
KBig news everyone. Wait for tomorrow https://www.reddit.com/r/StableDiffusion/comments/13q4ku4/nvidia_2x_performance_improvement_for_stable/
reddit
523 votes and 123 comments so far on Reddit
AIts already 24th May for me
H@Furkan Gözükara SECourses i trained a person model , when i using text2imge, it seems that the person face is good before 18-20steps,but reached 25 steps, the face is distorted a little ;the steps is 25, cfg is 7.5
Hit didnot works when i change the total steps 18
EWow.. Thats really huge improvements for performance
Mhey guys, I'm training a model on an architectural style. I have a decent number of training images, about 500. Given the general guide of 100 steps per training image for styles, that would mean 50000 steps, right?
MDo more training images work better for this? Or are there diminishing returns
Fso why is this happening to a lot of people? xD
AI am using runpod for the first time and get this error when I start training. Does anyone know how to fix it?
ANever mind, must have forgotten to restart it after changing the torch version, its working now.
Fwhat torch version you using?
GDoes anyone know how to make the Stable diffusion output look more realistic? This output looks like someone applied a cookie-cutter filter
Bhi friends! can you use controlnet with custom checkpoints?
ADowngraded it from 2 to 1.13.1 because 2 doesn't work properly on Linux
PYouTubeSECourses
Discord : https://bit.ly/SECoursesDiscord. Torch 2 / Pytorch 2 is now supported along with new DreamBooth Automatic1111 Web UI extension. 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
In this video, I am explaining how you can man...
https://www.patreon.com/SECoursesIn this video, I am explaining how you can man...
PSo are these steps still required if using the dev version of A1111?
PFor the install of Dreambooth?
POr can you just install the Dreambooth extension from the Extensions tab if you are using the dev version of Automatic1111 and bypass the venv folder updates that are in the above video?
RAnybody know if dreambooth extension on auto 1111 takes into account subfolders in the dataset directory?
Fof course
Bdo you happen to have a tutorial @felixsanz ? i tried a few things, but nothing is resulting in good outcomes yet.
BGitHub
News This post is out-of-date and obsolete. Please directly use Mikubill' A1111 Webui Plugin to control any SD 1.X models. No transfer is needed. Results are a bit better than the ones in this ...
Bwhich originally recommends to use: tool_transfer_control.py
Bbut now they just suggest using https://github.com/Mikubill/sd-webui-controlnet
Bi wanna implement this in code instead of the UI
Fyou don't need a tutorial other than just know how to use controlnet. the checkpoint that you use doesn't matter
Fyou don't read much, do you?
Fthe first like of your link says: "This post is out-of-date and obsolete. "
B
yeah, thats why i said originallyBthe repo was initially recommending that approach, and now it changed
Bbut i dont want to use the UI