Software Engineering Courses (SECourses)•3y ago

I'm trying to figure out why when i train my model, it's doing everything except what I want. i crea

I'm trying to figure out why when i train my model, it's doing everything except what I want. i created captions for each picture like this

Zono50OP•3/20/23, 10:19 AM

yet i'm getting images like this

sample_8016-0bdd9b21363dafdac629e9dd4219378c8ad12baa9.png

ZZono50 I'm trying to figure out why when i train my model, it's doing everything except...

Furkan Gözükara SECourses•3/20/23, 10:21 AM

I never shown such instance token setup

Furkan Gözükara SECourses•3/20/23, 10:21 AM

wait my new video I explained captions

Dincman•3/20/23, 10:21 AM

thank you for answering.

DDincman thank you for answering.

Furkan Gözükara SECourses•3/20/23, 10:21 AM

what gpu you have in your pc?

Zono50OP•3/20/23, 10:21 AM

i used the tagger extension

Furkan Gözükara SECourses•3/20/23, 10:22 AM

fixing subtitles of the new video

Furkan Gözükara SECourses•3/20/23, 10:22 AM

Zono50OP•3/20/23, 10:23 AM

i assume by 50 epochs if i'm not seeing my subject, to stop and reassess

Zono50OP•3/20/23, 10:29 AM

you know me, my signature is breaking it

ZZono50 you know me, my signature is breaking it

Furkan Gözükara SECourses•3/20/23, 10:38 AM

must be a config error i think

Furkan Gözükara SECourses•3/20/23, 10:39 AM

i noticed an incorrect information in the video now reexporting :/

Zono50OP•3/20/23, 10:45 AM

quick question, if i provide my own classification dataset directory, can i set class images per instance image to 0 under class image generation?

ZZono50 quick question, if i provide my own classification dataset directory, can i set ...

Furkan Gözükara SECourses•3/20/23, 10:45 AM

yes

Furkan Gözükara SECourses•3/20/23, 10:45 AM

but it uses 0 then

Furkan Gözükara SECourses•3/20/23, 10:45 AM

calculate how many you can use

Furkan Gözükara SECourses•3/20/23, 10:45 AM

enter that

Zono50OP•3/20/23, 10:45 AM

Furkan Gözükara SECourses•3/20/23, 10:45 AM

so it wont generate new

Zono50OP•3/20/23, 10:59 AM

what's the video your uploading about?

Dkleo•3/20/23, 10:59 AM

@Zono50 you're getting these random images, because you sample image prompt is empty

Zono50OP•3/20/23, 11:00 AM

it says in the box, leave blank to use instance prompt, so it should be using photo of ohwx person, award winning photography I thought

Zono50OP•3/20/23, 11:03 AM

so technicallly i could do "a photo of ohwx person" for my instance prompt, and person, as my class prompt, then fill those fields as such

Instance Prompt:
[filewords]

class prompt:
[filewords]

Sample image prompt:
[filewords]

ZZono50 what's the video your uploading about?

Furkan Gözükara SECourses•3/20/23, 11:03 AM

style teaching + yourself

Dkleo•3/20/23, 11:05 AM

" leave blank to use instance prompt", i did not notice this part before you pointed it out. i usually do [filewords], because this image you're getting does look like a generation with no prompt,

Zono50OP•3/20/23, 11:08 AM

if i have 22 instance images, and a folder with like 1500 class images, what would be a good number to set for class images per instance image

Dkleo•3/20/23, 11:11 AM

@Furkan Gözükara SECourses has a video where he tested several numbers of class images per image, i forgot in what video exactly, maybe he can link it for you

ZZono50 if i have 22 instance images, and a folder with like 1500 class images, what wou...

Sexy skeleton•3/20/23, 11:11 AM

i had same problem, i'm solve it just by thaining not from dreambooth, i train it on dreambooth lora (kohya)

Zono50OP•3/20/23, 11:11 AM

kohya is broken in linux atm unfortunately

Zono50OP•3/20/23, 11:11 AM

it won't even start training

Sexy skeleton•3/20/23, 11:12 AM

that's sad

Zono50OP•3/20/23, 11:12 AM

for real

Dkleo•3/20/23, 11:12 AM

here it is : https://www.youtube.com/watch?v=Tb4IYIYm4os

YouTubeSECourses - Software Engineering Courses

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...

Zono50OP•3/20/23, 11:13 AM

thank you!

ZZono50 for real

Sexy skeleton•3/20/23, 11:21 AM

umm i'm not sure that it will be work, but you can try that as well, i guess. Just download windows sandbox and install on it (virtual machine) all your programs that not works on linux

Zono50OP•3/20/23, 12:01 PM

how do i save my images to a folder in txt2img if i do generate forever

Zono50OP•3/20/23, 12:02 PM

i want to create a folder of regularization images for woman

Zono50OP•3/20/23, 12:04 PM

nvmd found it

Garreth210•3/20/23, 12:11 PM

Hey! So I got stable-Diffusion auto1111 running and the dreambooth extention installed both on my local and on runpod. However on the local machine I get the following message when trying to create a model

https://gdl.space/lajekudili.coffeescript

https://gdl.space/lajekudili.coffeescript

Any help would be amazing. I am also following the zero the hero tutorial

GGarreth210 Hey! So I got stable-Diffusion auto1111 running and the dreambooth extention ins...

Furkan Gözükara SECourses•3/20/23, 12:20 PM

what is the message?

Furkan Gözükara SECourses•3/20/23, 12:20 PM

if you dont have at least 12 gb vram add this to command line arguments

Furkan Gözükara SECourses•3/20/23, 12:20 PM

--ckptfix

Garreth210•3/20/23, 12:22 PM

let me give that a try

FFurkan Gözükara SECourses what is the message?

Garreth210•3/20/23, 12:28 PM

Success

Loading model from checkpoint.
Loading ckpt...
Pred and size are epsilon and 512, using config: C:\stable-diffusion-webui\extensions\sd_dreambooth_extension\dreambooth\..\configs\v1-training-default.yaml
v1 model loaded.
Trying to load: C:\stable-diffusion-webui\extensions\sd_dreambooth_extension\dreambooth\..\configs\v1-training-default.yaml
Converting unet...
Converting vae...
Converting text encoder...
Saving text_encoder
Saving tokenizer
Checkpoint successfully extracted to C:\stable-diffusion-webui\models\dreambooth\GarrethD_VamphuntD-model\working
Restored system models.
Duration: 00:00:23
Model dir set to: C:\stable-diffusion-webui\models\dreambooth\GarrethD_VamphuntD-model
Using CPU for extraction.
Loading model from checkpoint.
Loading ckpt...

Loading model from checkpoint.
Loading ckpt...
Pred and size are epsilon and 512, using config: C:\stable-diffusion-webui\extensions\sd_dreambooth_extension\dreambooth\..\configs\v1-training-default.yaml
v1 model loaded.
Trying to load: C:\stable-diffusion-webui\extensions\sd_dreambooth_extension\dreambooth\..\configs\v1-training-default.yaml
Converting unet...
Converting vae...
Converting text encoder...
Saving text_encoder
Saving tokenizer
Checkpoint successfully extracted to C:\stable-diffusion-webui\models\dreambooth\GarrethD_VamphuntD-model\working
Restored system models.
Duration: 00:00:23
Model dir set to: C:\stable-diffusion-webui\models\dreambooth\GarrethD_VamphuntD-model
Using CPU for extraction.
Loading model from checkpoint.
Loading ckpt...

Zono50OP•3/20/23, 12:30 PM

i'm a generate 2400 images of woman, then set my num class instances of image to 50 and see how that does me

Garreth210•3/20/23, 12:39 PM

I just want to understand that, I am running a 3070Ti and I have dreambooth running now after adding the --ckptfix arag. Do I continue as normal or Do I need to enable Lora?

Zono50OP•3/20/23, 12:44 PM

What is the relationship between classification dataset directory and Class images per instance image

Furkan Gözükara SECourses•3/20/23, 12:45 PM

which one is best to use in thumbnail

GGarreth210 I just want to understand that, I am running a 3070Ti and I have dreambooth runn...

Furkan Gözükara SECourses•3/20/23, 12:46 PM

continue as normal after that. that is for generate training model

FFurkan Gözükara SECourses continue as normal after that. that is for generate training model

Garreth210•3/20/23, 12:46 PM

perfect.

Loading model from checkpoint. Loading ckpt... Pred and size are epsilon and 512, using config: C:\stable-diffusion-webui\extensions\sd_dreambooth_extension\dreambooth\..\configs\v1-training-default.yaml v1 model loaded. Trying to load: C:\stable-diffusion-webui\extensions\sd_dreambooth_extension\dreambooth\..\configs\v1-training-default.yaml Converting unet... Converting vae... Converting text encoder... Saving text_encoder Saving tokenizer Checkpoint successfully extracted to C:\stable-diffusion-webui\models\dreambooth\GarrethD_VamphuntD-model\working Restored system models. Duration: 00:00:23 Model dir set to: C:\stable-diffusion-webui\models\dreambooth\GarrethD_VamphuntD-model Using CPU for extraction. Loading model from checkpoint. Loading ckpt...

I'm trying to figure out why when i train my model, it's doing everything except what I want. i crea

Similar Threads