Software Engineering Courses (SECourses)•2y ago

Testing OneTrainer masked Stable Diffusion XL (SDXL) training results + OneTrainer vs Kohya trainer

Testing OneTrainer masked Stable Diffusion XL (SDXL) training results + OneTrainer vs Kohya trainer results with 4 GPUs at the same time. Total 10 model is being tested with Automatic1111 x/y/z checkpoint comparison.

FFurkan Gözükara SECourses https://github.com/FurkanGozukara/Stable-Diffusion/blob/main/Tutorials/OneTraine...

ch1253•4/3/24, 12:54 AM

I could see the youtube video for windows but for ubuntu in MassedCompute after updating according to the above reference starting the OneTrainer on Ubuntu, do we have to install all the same things same as in windows? Note: I have selected the same vms as given in the image. Thank you!

Cch1253 I could see the youtube video for windows but for ubuntu in MassedCompute after ...

Furkan Gözükara SECoursesOP•4/3/24, 12:55 AM

i am working on video

Furkan Gözükara SECoursesOP•4/3/24, 12:55 AM

but if you follow readme

Furkan Gözükara SECoursesOP•4/3/24, 12:55 AM

you can do i think

ch1253•4/3/24, 1:01 AM

Sure, May I know probable time of the video being published?

Cch1253 Sure, May I know probable time of the video being published?

Furkan Gözükara SECoursesOP•4/3/24, 1:03 AM

hopefully this week

Furkan Gözükara SECoursesOP•4/3/24, 1:03 AM

i am doing final tests

AIGambino•4/3/24, 1:22 AM

woooo! custom train hopefully it works this time haha

FFurkan Gözükara SECourses Testing OneTrainer masked Stable Diffusion XL (SDXL) training results + OneTrain...

metta9403•4/3/24, 3:52 AM

This is amazing

Furkan Gözükara SECoursesOP•4/3/24, 4:30 AM

new comparison

Furkan Gözükara SECoursesOP•4/3/24, 4:30 AM

https://www.reddit.com/r/StableDiffusion/comments/1bujpn3/onetrainer_fine_tuning_vs_kohya_ss_dreambooth/

From the StableDiffusion community on Reddit: OneTrainer Fine Tunin...

Explore this post and more from the StableDiffusion community

Furkan Gözükara SECoursesOP•4/3/24, 4:41 AM

i got to sleep good day / night

Tuanny•4/3/24, 6:55 AM

using Supir to 8x a 800x1700 image on my rtx 3080 with 10vram taking about 2.5 hrs to upscale
if i upgrade to like a 4090 with 24gb will it take like 20-30 mins to upscale to 8x?

FFurkan Gözükara SECourses i would compare supir low text cfg upscaled

kareem9000•4/3/24, 10:19 AM

https://www.patreon.com/posts/supir-full-1-new-99349514

Patreon

SUPIR Full Tutorial + 1 Click 12GB VRAM Windows & RunPod / Linux In...

Get more from SECourses: Tutorials, Guides, Resources, Training, MidJourney, Voice Clone, TTS, ChatGPT, GPT, LLM, Scripts by Furkan Gözükara on Patreon

kareem9000•4/3/24, 10:19 AM

so technically i would try to run the dataset by supir before i start training on dreambooth

ArtChicken•4/3/24, 10:20 AM

Is it possible to extract a lora from finetuned model with Onetrainer?

joshtech•4/3/24, 10:33 AM

@Dr. Furkan Gözükara Thanks to your great guides I now have a really great dreambooth model that reproduces the face correctly in most situations

. However I still have big problems with hands

. I need to generate a lot of images of the person and I dont have time to manually fix certain images. That's why I have to sort out about 80% of ALL images because the hands are misformed. Adetailer does not do a good job at fixing hands too. Do you have any recommendation for a workflow that either avoids hands in the image or at least fixes them sometimes? I cannot use "hands in pocket" because the images are NSFW mostly... Thanks for your ideas. You do an amazing job, really!

Jjoshtech @Dr. Furkan Gözükara Thanks to your great guides I now have a really great dream...

ArtChicken•4/3/24, 11:53 AM

try to use embeddings: they don't do miracles, but sometime prevent big mistakes (that adetailer can fix more easily); try to search specific embeddings on civitai and play around...

TTuanny using Supir to 8x a 800x1700 image on my rtx 3080 with 10vram taking about 2.5 ...

Furkan Gözükara SECoursesOP•4/3/24, 3:06 PM

it would take few minutes

Furkan Gözükara SECoursesOP•4/3/24, 3:07 PM

unless it uses shared vram

Furkan Gözükara SECoursesOP•4/3/24, 3:07 PM

i can test it

Furkan Gözükara SECoursesOP•4/3/24, 3:07 PM

tell me your final resolution that you see

Kkareem9000 so technically i would try to run the dataset by supir before i start training o...

Furkan Gözükara SECoursesOP•4/3/24, 3:07 PM

yes

Furkan Gözükara SECoursesOP•4/3/24, 3:07 PM

but i would compare

Furkan Gözükara SECoursesOP•4/3/24, 3:07 PM

both original and upscaled

AArtChicken Is it possible to extract a lora from finetuned model with Onetrainer?

Furkan Gözükara SECoursesOP•4/3/24, 3:07 PM

yes but i prefer kohya better

Jjoshtech @Dr. Furkan Gözükara Thanks to your great guides I now have a really great dream...

Furkan Gözükara SECoursesOP•4/3/24, 3:08 PM

hands issue coming from the base model

Furkan Gözükara SECoursesOP•4/3/24, 3:08 PM

and thanks

rafraf•4/3/24, 3:19 PM

@Dr. Furkan Gözükara in kohya, can "Epoch" be set to "0" or empty if I set Max Steps to 10000? I am having issue when i set sample every 100 steps but leave Sample every "0" epochs. Should I leave sample every x epochs? I'm struggling finding what is the accepted parameter

rafraf•4/3/24, 3:26 PM

standard_1.json4.4KB

virginia•4/3/24, 6:26 PM

@Dr. Furkan Gözükara can webui still be used?

FFurkan Gözükara SECourses both original and upscaled

kareem9000•4/3/24, 9:20 PM

Can you explain what this means? do you mean do an original supir and then an upscale supir?

AIGambino•4/4/24, 12:12 AM

one trainer won't train what am I missing