Software Engineering Courses (SECourses)•3y ago

speeds up training alot on nvidia gpus, if you are cpu training disregard

Spydah•5/8/23, 9:32 PM

WRT Kohya on runpod... https://github.com/bmaltais/kohya_ss/discussions/379

GitHub

runpod.io kohya_ss directions (in thread) · bmaltais kohya_ss · Dis...

I had some trouble with the other linux ports (& the kohya_ss-linux that runpod has as a template) instead you can use the latest bmaltais/kohya_ss fork: deploy their existing RunPod Stable Dif...

TTheRustingDeadAdmin speeds up training alot on nvidia gpus, if you are cpu training disregard

Dagbarox•5/8/23, 9:32 PM

Thanks for the help, I haven't tried training yet. Maybe I'll try soon im not sure

FFurkan Gözükara SECourses and i havent looked for runpod solution

poeticoncept•5/8/23, 9:37 PM

thanks for replying sir, i'm just trying to train lora on runpod

. I'm "Miguel Arce" you recommend me to try dreambooth insteed with runpod, i will give it a try.

Zono50•5/8/23, 9:52 PM

is there any kind of opensource reface app?

Zono50•5/8/23, 9:52 PM

i would love to put my face on hulk hogan as he walks to the ring

Furkan Gözükara SECourses•5/8/23, 9:56 PM

@Kooper now i will use this repo to auto install

Furkan Gözükara SECourses•5/8/23, 9:56 PM

https://github.com/rogerb831/runpod-dreambooth-easy

GitHub

GitHub - rogerb831/runpod-dreambooth-easy

Contribute to rogerb831/runpod-dreambooth-easy development by creating an account on GitHub.

Furkan Gözükara SECourses•5/8/23, 9:56 PM

this is in pinned comments

DDagbarox Guys I'm very new to Stable Diffusion, do I need cuDNN or is it not necessary? I...

Furkan Gözükara SECourses•5/8/23, 9:58 PM

11.8 good

SSpydah WRT Kohya on runpod... https://github.com/bmaltais/kohya_ss/discussions/379

Furkan Gözükara SECourses•5/8/23, 9:59 PM

thanks i will check

Ppoeticoncept thanks for replying sir, i'm just trying to train lora on runpod 😢 . I'm "Migue...

Furkan Gözükara SECourses•5/8/23, 9:59 PM

dreambooth better

Furkan Gözükara SECourses•5/8/23, 9:59 PM

i compared dreambooth and lora and i think dreambooth better

FFurkan Gözükara SECourses 11.8 good

Dagbarox•5/8/23, 10:05 PM

Thanks hocam, I've learnt a lot from your videos, you're the best teacher especially for newcomers ^^ adamsınız

DDagbarox Thanks hocam, I've learnt a lot from your videos, you're the best teacher especi...

Furkan Gözükara SECourses•5/8/23, 10:06 PM

thank you so much

Furkan Gözükara SECourses•5/8/23, 10:06 PM

@Spydah i just did a fresh install

Furkan Gözükara SECourses•5/8/23, 10:06 PM

and this time it didnt work

Furkan Gözükara SECourses•5/8/23, 10:06 PM

very weird

Spydah•5/8/23, 10:06 PM

I'll do one right now

Spydah•5/8/23, 10:09 PM

just worked flawlessly for me on a fresh 1 x RTX A6000, 12 vCPU 62 GB RAM, runpod/stable-diffusion:web-automatic-5.0.0

SSpydah just worked flawlessly for me on a fresh 1 x RTX A6000, 12 vCPU 62 GB RAM, runpo...

Furkan Gözükara SECourses•5/8/23, 10:12 PM

very weird

Furkan Gözükara SECourses•5/8/23, 10:12 PM

i did a new manual install it worked

Furkan Gözükara SECourses•5/8/23, 10:12 PM

i wonder if some pods having initial install problem

Spydah•5/8/23, 10:13 PM

Possibly. I'd need to see the output from the console to see what might have failed.

Furkan Gözükara SECourses•5/8/23, 10:18 PM

dreambooth load was failing interestingly config bla bla error

Furkan Gözükara SECourses•5/8/23, 10:18 PM

this morning i had successful install

Spydah•5/8/23, 10:19 PM

Must have been a fluke. It doesn't really check for errors and they could easily get buried in the output.

Furkan Gözükara SECourses•5/8/23, 10:20 PM

man the scripts i prepared for voice clone now awesome

Furkan Gözükara SECourses•5/8/23, 10:20 PM

you can see from this file already :d https://github.com/FurkanGozukara/Stable-Diffusion/blob/main/Tutorials/Deep-Voice-Clone-Tutorial-Tortoise-TTS.md

GitHub

Stable-Diffusion/Deep-Voice-Clone-Tutorial-Tortoise-TTS.md at main ...

Best Stable Diffusion and AI Tutorials, Guides, News, Tips and Tricks - Stable-Diffusion/Deep-Voice-Clone-Tutorial-Tortoise-TTS.md at main · FurkanGozukara/Stable-Diffusion

Individual•5/8/23, 10:21 PM

Someone know how to relaunch the Runpod Camenduru Stable Diffusion webui?

FFurkan Gözükara SECourses i compared dreambooth and lora and i think dreambooth better

TheRustingDeadAdminOP•5/8/23, 10:22 PM

a1111 dreambooth better than koyha lora?

Spydah•5/8/23, 10:24 PM

Especially on windows. None of the file selection widget work on linux with Koyha

TTheRustingDeadAdmin a1111 dreambooth better than koyha lora?

Furkan Gözükara SECourses•5/8/23, 10:29 PM

Yep

Furkan Gözükara SECourses•5/8/23, 10:29 PM

I think lora better only if you have 8 gb vram

IIndividual Someone know how to relaunch the Runpod Camenduru Stable Diffusion webui?

Furkan Gözükara SECourses•5/8/23, 10:29 PM

I didn't test it

Furkan Gözükara SECourses•5/8/23, 10:30 PM

My tutorial still works

FFurkan Gözükara SECourses I didn't test it

Individual•5/8/23, 10:32 PM

Can't really as there is no relauncher.py so I was not able to input "python relauncher.py" into the terminal. I'll just keep on trying.

Individual•5/8/23, 10:41 PM

ok nvm I got it working now

FFurkan Gözükara SECourses I think lora better only if you have 8 gb vram

TheRustingDeadAdminOP•5/8/23, 11:00 PM

thats me! thank you dr

Furkan Gözükara SECourses•5/8/23, 11:09 PM

finally finished recording of the tutorial phew it took a lot

Furkan Gözükara SECourses•5/9/23, 12:20 AM

i even explained loss value batch size step count

Furkan Gözükara SECourses•5/9/23, 12:20 AM

there is no info about these anywhere else :d

Furkan Gözükara SECourses•5/9/23, 12:22 AM

post processing of this tutorial will take a lot of time as well

Furkan Gözükara SECourses•5/9/23, 12:22 AM

i hope i can get some new patreon supporters from this video :d

Furkan Gözükara SECourses•5/9/23, 1:30 AM

@mtnmecca this is the correct place

Furkan Gözükara SECourses•5/9/23, 1:30 AM

yes the script may not detect face in some images that is expected

Furkan Gözükara SECourses•5/9/23, 1:30 AM

the important thing is

Furkan Gözükara SECourses•5/9/23, 1:30 AM

all images in your original folder should be face detectable

mtnmecca•5/9/23, 2:06 AM

Is there a size recommendation like 512x512 or distance from camera that is better? I used 5 pics and it only recognized 1. If there is more info on Deepface I can research tomorrow

mtnmecca•5/9/23, 2:13 AM

I’ll comment on your latest high quality image video tomorrow but I wanted to tell you that I updated to newer versions of xformers, torch and others and my image generation quality (especially skin texture) went way up. I have both old and new versions installed so the comparison is really good. I didn’t know the quality could improve this much through updating SD component code. I have spent so much time trying to improve training but good of improvement was also achieved by newer code updates..

weilu•5/9/23, 5:19 AM

@Furkan Gözükara SECourses Do we need to use blip or tagger to label abstract styles when training with Dreambooth?

speeds up training alot on nvidia gpus, if you are cpu training disregard

Similar Threads

Similar Threads

Similar Threads