Software Engineering Courses (SECourses)•3y ago

@Dr. Furkan Gözükara I am using 150 training steps per image and have 30 images in my dataset. I am

@Dr. Furkan Gözükara I am using 150 training steps per image and have 30 images in my dataset. I am not using any classification images. Do you have any advice why does my train model look different from my dataset?

LEGACIESOP•10/8/23, 10:37 AM

Sandstormleader•10/8/23, 10:44 AM

what tokens are you using?

Sandstormleader•10/8/23, 10:45 AM

if you just gave it a standard name the model could already have some info on that for example if you are using "olivia" vs something like "ohwx"

LEGACIESOP•10/8/23, 10:57 AM

Sandstormleader•10/8/23, 10:57 AM

i don't even put woman in my instance prompt personally

Sandstormleader•10/8/23, 10:59 AM

not sure if that matters but @Dr. Furkan Gözükara usually suggest class images when he does it that way or when he trains in general

LEGACIESOP•10/8/23, 11:01 AM

i will try to train another one without the word woman

LEGACIESOP•10/8/23, 11:02 AM

how many class images do i need for 30images in dataset?

Sandstormleader•10/8/23, 11:03 AM

that should be plenty from my experience

Sandstormleader•10/8/23, 11:04 AM

i've done some with 20 and others with 200

Sandstormleader•10/8/23, 11:05 AM

over training and bad quality have usually been my only issues with not using class images, it's pretty much always looked like my dataset

LEGACIESOP•10/8/23, 11:09 AM

so you always used class images when doing training?

Sandstormleader•10/8/23, 11:09 AM

not me no,

Sandstormleader•10/8/23, 11:09 AM

i also don't downsize my images or anything like that

Sandstormleader•10/8/23, 11:10 AM

i read somewhere on reddit that it actually does all that for you in theory. idk anything official but i know it's been working for me

LEGACIESOP•10/8/23, 11:10 AM

training model is so difficult XD

Sandstormleader•10/8/23, 11:10 AM

are you using 1.5 or sdxl?

LEGACIESOP•10/8/23, 11:11 AM

1.5

Sandstormleader•10/8/23, 11:11 AM

is it a lora or ckpt?

LEGACIESOP•10/8/23, 11:11 AM

checkpoint

Sandstormleader•10/8/23, 11:11 AM

ok ya that's same as me

LEGACIESOP•10/8/23, 11:12 AM

i dont understand why does my model look different from my dataset

Sandstormleader•10/8/23, 11:13 AM

did you double check the obvious stuff lol i've messed up in that way plenty of times

Sandstormleader•10/8/23, 11:13 AM

no lora's on, correct ckpt selected, correct folder linked to dataset input,

LEGACIESOP•10/8/23, 11:14 AM

yes these setting are good

LEGACIESOP•10/8/23, 11:14 AM

I will test another one without class images

LEGACIESOP•10/8/23, 11:15 AM

if if it does not look good then i will test it with class images

LEGACIESOP•10/8/23, 11:15 AM

btw what gpu are you using?

Sandstormleader•10/8/23, 11:39 AM

RTX 4090

Sandstormleader•10/8/23, 11:42 AM

should work with other cards tho too since i leave on xformers and 8bit adam

SSandstormleader @Stefanoid9343 from my experience the double the double people thing really depe...

Stefanoid9343•10/8/23, 12:09 PM

Hi! I’m using real vision model the same as the youtube video

SStefanoid9343 Hi @Dr. Furkan Gözükara I follow your runpod tutorial for creating portrait phot...

Furkan Gözükara SECourses•10/8/23, 12:29 PM

which base model you used?

Furkan Gözükara SECourses•10/8/23, 12:29 PM

it affects huge

Eelie Hi, could someone plz point me to the right resources for incorporating two safe...

Furkan Gözükara SECourses•10/8/23, 12:30 PM

you can use checkpoint merger of automatic1111 web ui and experiment with it

Furkan Gözükara SECourses•10/8/23, 12:31 PM

Furkan Gözükara SECourses•10/8/23, 12:31 PM

https://youtu.be/s25hcW4zq4M

YouTubeSECourses

How to Inject Your Trained Subject e.g. Your Face Into Any Custom S...

Our Discord : https://discord.gg/HbqgGaZVmr. In this video, I am explaining how to extract the taught subject information from your trained model and inject into any new custom Stable Diffusion model. 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.patreo...

SStefanoid9343 should I add negative prompts before start the training? also I’m getting double...

Furkan Gözükara SECourses•10/8/23, 12:31 PM

they are used during inference. not during training

Furkan Gözükara SECourses•10/8/23, 12:31 PM

1024x1024 better for sdxl

Sandstormleader•10/8/23, 12:32 PM

@Dr. Furkan Gözükara any advice for best batch upscaling?

FFurkan Gözükara SECourses you can use checkpoint merger of automatic1111 web ui and experiment with it

elie•10/8/23, 12:34 PM

I'm just trying to learn how to do this on python instead of using a ui. Are there any huggingface loaders that allow this or do I need to look at kohyas scripts?

FFurkan Gözükara SECourses they are used during inference. not during training

Stefanoid9343•10/8/23, 12:42 PM

Should I use 1024x1024 instead? should I add negative prompts before start the training ?

SSandstormleader @Dr. Furkan Gözükara any advice for best batch upscaling?

Furkan Gözükara SECourses•10/8/23, 12:44 PM

there is controlnet upscale with hd upscaler maybe that one

Furkan Gözükara SECourses•10/8/23, 12:44 PM

https://youtu.be/3E5fhFQUVLo

YouTubeSECourses

Zero to Hero ControlNet Tutorial: Stable Diffusion Web UI Extension...

I show how to install Automatic1111 Web UI & ControlNet extension installation from scratch in this video. Moreover I show how to make amazing QR codes and inpainting and out painting of ControlNet which are very similar to Photoshop generative fill and Midjourney zoom out. Furthermore, I explain and show what are Canny, Depth, Normal, OpenPose...