Software Engineering Courses (SECourses)•3y ago

every gude in the planet says one should use OHWX MAN! removing man.. well in my case girl solved ev

every gude in the planet says one should use OHWX MAN! removing man.. well in my case girl solved everything.. I've been strugling with this for MONTHS! @Furkan Gözükara SECourses hey make a video about that man..

felixsanz•5/23/23, 10:16 PM

well that's obvious xD i don't use woman in my prompting, because you train that word with the regularization images

felixsanz•5/23/23, 10:18 PM

but i don't understand why when i was a noob i trained perfect models all the time, and now i tried everything and all models suck

Ffelixsanz well that's obvious xD i don't use woman in my prompting, because you train that...

AlexGilSegOP•5/23/23, 10:21 PM

how the hell is that obvious?? Every guide on youtube and what ever I find on reddit says to use sks person or ohwx man etc etc...

felixsanz•5/23/23, 10:22 PM

but you're training random class images on that word!

AlexGilSegOP•5/23/23, 10:23 PM

I knooooooooow! It's insane!!!

AlexGilSegOP•5/23/23, 10:23 PM

but for all my other models.. of me my wife my other daughter it has worked fine.. here no for some reason..

AlexGilSegOP•5/23/23, 10:25 PM

I feel like an idiot.. I've made at least 100 models.. at leeast.. and now I find this out lol

felixsanz•5/23/23, 10:36 PM

lol dude, it was easy to just try another prompt xd

Preacher•5/23/23, 10:55 PM

Missing model directory, removing model: C:\Ai\stable-diffusion-webui\models\dreambooth\ft7\working\text_encoder

How do i fix this

Ffelixsanz lol dude, it was easy to just try another prompt xd

AlexGilSegOP•5/23/23, 10:59 PM

again.. ALL bloody tutorials say it wrong... also I've made COUNTLESS mosts on reddit.. not one suggestion to remove class token from my prompt.. makes me so mad..

Ffelixsanz but i have trained some models 2-3 months ago and pictures are awesome with just...

Furkan Gözükara SECourses•5/23/23, 11:14 PM

must be config error

FFurkan Gözükara SECourses must be config error

felixsanz•5/23/23, 11:14 PM

but what config? the other models works fine

felixsanz•5/23/23, 11:15 PM

i switch model and it runs fine

AAlexGilSeg every gude in the planet says one should use OHWX MAN! removing man.. well in my...

Furkan Gözükara SECourses•5/23/23, 11:15 PM

you removed from prompting or training

Furkan Gözükara SECourses•5/23/23, 11:15 PM

i test that man class effect

Furkan Gözükara SECourses•5/23/23, 11:15 PM

class images performed better

felixsanz•5/23/23, 11:15 PM

and if you mean training config..... no matter what I do. even if I replicate your videos 100%

Ffelixsanz and if you mean training config..... no matter what I do. even if I replicate yo...

Furkan Gözükara SECourses•5/23/23, 11:16 PM

i mean training configuration. can you try with training a man?

Furkan Gözükara SECourses•5/23/23, 11:16 PM

we have shared training dataset in this video : https://youtu.be/dNOpWt-epdQ

YouTubeSECourses

How To Do Stable Diffusion Textual Inversion (TI) / Text Embeddings...

Our Discord : https://discord.gg/HbqgGaZVmr. Grand Master tutorial for Textual Inversion / Text Embeddings. 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 Guide...

Original message was deleted

felixsanz•5/23/23, 11:17 PM

I can, but remember i trained this model perfectly fine 2-3 months ago

Furkan Gözükara SECourses•5/23/23, 11:18 PM

and i tested using class images effect in this video : https://youtu.be/Tb4IYIYm4os

YouTubeSECourses

Automatic1111 Stable Diffusion DreamBooth Guide: Optimal Classifica...

Sign up RunPod: https://bit.ly/RunPodIO. Our Discord : https://discord.gg/HbqgGaZVmr. New best training settings for DreamBooth training in Automatic1111 Web UI. 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 #StableDif...

Original message was deleted

felixsanz•5/23/23, 11:18 PM

why one woman looks like that, and the other like this? xD

Furkan Gözükara SECourses•5/23/23, 11:18 PM

i can train your training images if you wish

Furkan Gözükara SECourses•5/23/23, 11:18 PM

on my computer

felixsanz•5/23/23, 11:19 PM

i'm using runpod, like 20 times or so

felixsanz•5/23/23, 11:19 PM

i'm gonna try the man and report back

felixsanz•5/23/23, 11:20 PM

it's weird because it doesn't look like overtrained or undertrained

felixsanz•5/23/23, 11:20 PM

it is possible to view metadata from a model like when you view it from a picture?

felixsanz•5/23/23, 11:21 PM

like.. base model, seed, steps/epochs, learning rate.... etc

felixsanz•5/23/23, 11:21 PM

is all that viewable on a model?

felixsanz•5/23/23, 11:21 PM

probably not, but it's a good idea XD

Ffelixsanz it is possible to view metadata from a model like when you view it from a pictur...

Furkan Gözükara SECourses•5/23/23, 11:23 PM

i dont know if that is saved :d

felixsanz•5/23/23, 11:30 PM

i've see that you use "blue eyes" as the first negative prompt

felixsanz•5/23/23, 11:30 PM

i'm also always getting blue eyes

felixsanz•5/23/23, 11:30 PM

why do you think that happens?

Ffelixsanz why do you think that happens?

Furkan Gözükara SECourses•5/23/23, 11:37 PM

because custom models are biased

Furkan Gözükara SECourses•5/23/23, 11:37 PM

also some tokens are biased

Furkan Gözükara SECourses•5/23/23, 11:37 PM

like handsome or photoshoot :d

Furkan Gözükara SECourses•5/23/23, 11:38 PM

i am going to sleep hopefully see you tomorrow

Furkan Gözükara SECourses•5/23/23, 11:39 PM

i spent like 8 hours to install new cpu and motherboard

Furkan Gözükara SECourses•5/23/23, 11:39 PM

Furkan Gözükara SECourses•5/23/23, 11:39 PM

i plan to compare all optimizers in kohya lora

FFurkan Gözükara SECourses you removed from prompting or training

AlexGilSegOP•5/23/23, 11:40 PM

I 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..

felixsanz•5/23/23, 11:46 PM

i thinking that this crappy regularization images that i'm using are deteriorating the results

felixsanz•5/23/23, 11:46 PM

the woman class is re-trained with monsters

AAlexGilSeg I removed from prompting.. I still use the class images.. but removing girl from...

Furkan Gözükara SECourses•5/23/23, 11:50 PM

could be nice find

Ffelixsanz the woman class is re-trained with monsters

Furkan Gözükara SECourses•5/23/23, 11:50 PM

could be

felixsanz•5/23/23, 11:53 PM

i'm gonna try filewords + controlnet, someone told me that he's having a great result generating class images with controlnet depth map

felixsanz•5/23/23, 11:53 PM

so class images replicate the training images the most possible, but without the person face bias

AvidyaCM•5/24/23, 12:46 AM

if 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?

every gude in the planet says one should use OHWX MAN! removing man.. well in my case girl solved ev

Similar Threads