Stable Diffusion UI with Dreambooth, ControlNet, Deforum, roop extensions preinstalled & Kohya_ss | Runpod | Page 1

misty sierra Jun 12, 2023, 12:16 PM

#

question: how much VRAM does this template require?

modern quest Jun 12, 2023, 12:19 PM

#

Depends on what you want to do. For full Dreambooth, 24GB is recommended. You can use less if you train a LoRA.

modern quest Jun 27, 2023, 4:19 PM

#

This has just been updated to the latest Stable Diffusion Web UI 1.4.0 which was released about 10 hours ago.

I have also optimized the image so that it is almost 10GB smaller in size.

lusty lynx Jun 30, 2023, 2:13 AM

#

Awesome thank you very much Ashleyk ❤️ ❤️ ❤️

modern quest Jul 4, 2023, 4:18 PM

#

Template has been updated to include the roop face swap extension for A1111.

modern quest Jul 4, 2023, 4:41 PM

#

Stable Diffusion UI with Dreambooth, ControlNet, Deforum, roop extensions preinstalled & Kohya_ss

dreamy rapids Jul 9, 2023, 9:48 PM

#

Hi!, is this image also bloated? or is a pruned A1111?

modern quest Jul 13, 2023, 9:25 PM

#

dreamy rapids Hi!, is this image also bloated? or is a pruned A1111?

It includes a lot of extensions like Dreambooth, roop and also includes Kohya_ss so its pretty bloated as a result. If you are looking for something light-weight, then check out the Fast Stable Diffusion template.

main frigate Jul 15, 2023, 9:37 AM

#

I am very new here, we can at any public model right?

#

Add*

modern quest Jul 15, 2023, 9:40 AM

#

Yes you can download custom models into /workspace/stable-diffusion-webui/models/Stable-diffusion
You can either use wget to downlod from Huggingface, or else use download-model [download-link] /workspace/stable-diffusion-webui/models/Stable-diffusion to download models from CivitAI (both from terminal).

main frigate Jul 15, 2023, 10:11 AM

#

uhm this mean if I want to use this with my project I have to run the pod and use it as api or something like that?

modern quest Jul 15, 2023, 10:15 AM

#

If you are building some project and have not used A1111 Web UI for stable diffusion, it is good to first spin up a pod in GPU cloud to figure out how you are going to build your project. This template exposes the A1111 API which you can access by adding /api to the end of the URL in order to view the API documentation. However, once you are ready to start building the code for your project, I would recommend using RunPod Serverless for scalability. Generative Labs have some example videos on how to do that on their YouTube channel - https://www.youtube.com/@generativelabs/videos

YouTube

Generative Labs

main frigate Jul 15, 2023, 2:21 PM

#

oh thanks so much

main frigate Jul 15, 2023, 11:57 PM

#

Yeah I just deploy my first pod, let me try. I will try not ask too much question. 😅

main frigate Jul 16, 2023, 9:30 AM

#

thanks so much for your advise. I can install new model from civitai both model, lora, embedding. However, when I tried to install control net model from hugging face with this command wget https://huggingface.co/lllyasviel/ControlNet-v1-1/resolve/main/control_v11f1e_sd15_tile.pth /workspace/stable-diffusion-webui/extensions\sd-webui-controlnet\models .. It was wokring right like it will be fine. But It did not show up in controlnet model I can select

#

modern quest Jul 16, 2023, 9:36 AM

#

main frigate thanks so much for your advise. I can install new model from civitai both model,...

Your command looks incorrect, should be like this:

cd /workspace/stable-diffusion-webui/models/ControlNet
wget https://huggingface.co/lllyasviel/ControlNet-v1-1/resolve/main/control_v11p_sd15_openpose.pth
wget https://huggingface.co/lllyasviel/ControlNet-v1-1/resolve/main/control_v11p_sd15_canny.pth
wget https://huggingface.co/lllyasviel/ControlNet-v1-1/resolve/main/control_v11f1p_sd15_depth.pth
wget https://huggingface.co/lllyasviel/ControlNet-v1-1/resolve/main/control_v11p_sd15_inpaint.pth
wget https://huggingface.co/lllyasviel/ControlNet-v1-1/resolve/main/control_v11p_sd15_lineart.pth
wget https://huggingface.co/ioclab/ioc-controlnet/resolve/main/models/control_v1p_sd15_brightness.safetensors
wget https://huggingface.co/lllyasviel/ControlNet-v1-1/resolve/main/control_v11f1e_sd15_tile.pth

#

That should instlal all of them. If you are busy with QR codes, you probably want the last 2.

main frigate Jul 16, 2023, 9:50 AM

#

You are so kind.. Now I am going to install script https://github.com/Coyote-A/ultimate-upscale-for-automatic1111

GitHub

GitHub - Coyote-A/ultimate-upscale-for-automatic1111

Contribute to Coyote-A/ultimate-upscale-for-automatic1111 development by creating an account on GitHub.

#

I think it was install already but I cannot found in script.

#

modern quest Jul 16, 2023, 9:53 AM

#

If you want to install extensions, you need to edit /workspace/stable-diffusion-webui/webui-user.sh and remove --skip-install from the COMMANDLINE_ARGS arguments. It skips install by default to speed up the startup time.

main frigate Jul 16, 2023, 10:04 AM

#

ah I knew this concept that I can adjust argument before start. But I dont know where I can edit argument in runpod. I have to edit via web terminal?

modern quest Jul 16, 2023, 10:05 AM

#

web terminal, Jupyter terminal or SSH terminal

main frigate Jul 16, 2023, 10:09 AM

#

ah It is quite hard I have never used web terminal before

#

#

I can edit but I don't know how to save 😅

modern quest Jul 16, 2023, 10:10 AM

#

which editor is that? vi/vim?

#

just do this to save:

:wq!

main frigate Jul 16, 2023, 10:16 AM

#

I can type anywhere that is blank space? cause I cannot drag my cursor to the bottom.

modern quest Jul 16, 2023, 10:17 AM

#

if you type : it goes to the bottom

#

so yes just type it anywhere

main frigate Jul 16, 2023, 10:20 AM

#

Thanks so much I will try later. I think I can get it done soon. It is very amazing that I can use this service. Anyway, it is a bit pricy (for me) if I have to run this all the time waiting for user to make a request.

modern quest Jul 16, 2023, 10:21 AM

#

Yeah it is pricy for me too, so I use serverless beause serverless can also scale better

main frigate Jul 16, 2023, 10:21 AM

#

My current package

#

modern quest Jul 16, 2023, 10:21 AM

#

and you do't pay anything while there are no requests (unless you do something like set min workers)

#

Yeah, I was using A5000 to run Dreambooth and things and it costs even more 🙈

main frigate Jul 16, 2023, 10:41 AM

#

Ah you are not runpod staff?

modern quest Jul 16, 2023, 11:13 AM

#

No I'm not

main frigate Jul 16, 2023, 12:59 PM

#

Oh Thanks for your sharing again🫡🙏

modern quest Jul 25, 2023, 11:59 AM

#

Updated to Automatic1111 Web UI 1.5.0 and now has SDXL support 🚀

raven widget Jul 26, 2023, 4:37 AM

#

Hi Ashleyk, thanks for setting up this template and sharing it with everyone. I have a couple of questions about the usage.

Can I deploy it in Serverless, or it can only be used in GPU Cloud?
Assuming I can deploy it in Serverless, if I train a deambooth model, and want to use that model within automatic1111, do I have to mount the whole Docker image once again, or should it be available within the volume for me to use it?
About how many GB does the image size takes after installing everything?

Once again thanks!!!

modern quest Jul 26, 2023, 6:13 AM

#

Hi @raven widget .

No, this template is not designed to be used with serverless this is a GPU template. You can look at the GenerativeLabs videos for using A1111 with Serverless: https://www.youtube.com/@generativelabs/videos
You can bake custom models into your Serverless image as shown in the GenerativeLabs tutorials, or alternatively, you can store them on a Network Volume.
The image is around 24GB in size. The template sets 10GB container disk and 75GB persistent volume by default, but these can be customised upon deployment..

YouTube

Generative Labs

raven widget Jul 26, 2023, 6:21 AM

#

Thanks @modern quest ! I'll look more into them. I believe using the Network Volume to store the Dreambooth models and then calling them using a Serverless endpoint with automatic1111 will do the trick. Like a mix between what's in this video https://www.youtube.com/watch?v=b_5jkvMpd_k&ab_channel=GenerativeLabs and this one https://www.youtube.com/watch?v=gv6F9Vnd6io&ab_channel=GenerativeLabs.

YouTube

Generative Labs

RunPod Serverless, Network Volumes, Step-by-Step

In this in-depth tutorial, Billy from Generative Labs walks you through the steps to leverage Network Volumes with RunPod Serverless to create a Stable Diffusion API.

You'll learn how to set up a Network Volume, link it to your Serverless Endpoint, and we'll even guide you through downloading custom models from sources like civitai.com and Goo...

▶ Play video

YouTube

Generative Labs

Setting Up a Stable Diffusion API with Control Net using RunPod Ser...

In this step-by-step guide, we'll show you how to leverage the power of RunPod to create your own Stable Diffusion API with ControlNet enabled.

Here's what we'll cover in this tutorial:

Creating a Network Volume for robust model storage.
Installing Stable Diffusion and configuring it on the Network Volume.
Developing a Serverless Stable Diffus...

▶ Play video

#

I'll probably use your dockerfile as a guide to configure a .sh script to install a1111 along with dreambooth and kohya directly in a Network volume.

modern quest Jul 26, 2023, 6:55 AM

#

For installing Dreambooth, just make sure that the requirements versions for A1111 and Dreambooth extension match

modern quest Aug 1, 2023, 5:01 PM

#

Template has been updated so that you can add an environment variable called DOWNLOAD_SDXL with a value of 1 and it will automatically download the SDXL Base and Refiner models for you, as well as a styles.csv file that you can use to select various different SDXL styles from the dropdown by the Generate button. It also now includes the SDXL Refiner extension so that you can use the Base model and Refiner both from txt2img without having to send to img2img tab and switch out the models.

brazen river Aug 3, 2023, 10:54 AM

#

Hi man 😄 Using your template with only kohya atm. Is there any way to clone a template and edit it? Or only recreate yourself from scratch?

modern quest Aug 3, 2023, 10:57 AM

#

brazen river Hi man 😄 Using your template with only kohya atm. Is there any way to clone a t...

What is it that you would like to edit?

brazen river Aug 3, 2023, 11:07 AM

#

Well, i am new to all this, and my computer is to shit to run training etc. So I want an easy way to boot up, install all i need, download the model i want to try and then fetch my training data set. I started out just trying to setup pod step by step, installing kohya and it requirements, but i takes ages. I then found your template and i use that as a basis, download model manually and then runpodctl send the prepared folders from my own pc. But i still have to fiddle with apt-get update, closing stuff with fuser, trying to navigate jupityr lab. In the end i just want to create a serverless function that i can point to a dataset. That goal is far off though, for now i am only fiddeling a bit with it all. I took a look at your docker image and then i wondered why you can only clone official runpod templates in their console. So figured i might be missing something, and the other templates could be cloned/edited.

modern quest Aug 3, 2023, 11:10 AM

#

If the main thing you are interested in is Kohya_ss, I have a more light-weight template you can use for that:
https://runpod.io/gsc?template=51q837fywe&ref=2xxro4sy
And you can also get the code and modify it if you want:
https://github.com/ashleykleynhans/kohya-docker

GitHub

GitHub - ashleykleynhans/kohya-docker: Docker image for Kohya_ss We...

Docker image for Kohya_ss Web UI. Contribute to ashleykleynhans/kohya-docker development by creating an account on GitHub.

brazen river Aug 3, 2023, 11:11 AM

#

yeah, that is the one i am using now

modern quest Aug 3, 2023, 11:11 AM

#

Then you can just edit the Dockerfile and build your own docker image, template etc.

brazen river Aug 3, 2023, 11:11 AM

#

works good

#

what i basicly wanted was just to see this for the tempalte

#

template

#

#

to learn how it was setup

#

ahh, i can see them when i deploy the template

#

my bad

#

after choosing gpu

#

what confused me was just that in the manage->templates you can only see the official ones and duplicate those

#

sorry for the confusion

brazen river Aug 3, 2023, 1:35 PM

#

i will try out the full template later tonight 😄

sharp ledge Aug 4, 2023, 2:31 AM

#

can't add adetailer - it shows in the in extensions tab as installed but doesn't show up in txt2img or img2img - any ideas?

brazen river Aug 4, 2023, 4:05 PM

#

Not in relation to the previous poster, i haven't tried adetailer, just wanted to say that so far the template/container is working great for me

modern quest Aug 4, 2023, 4:11 PM

#

sharp ledge can't add adetailer - it shows in the in extensions tab as installed but doesn't...

Edit /workspace/stable-diffusion-webui/webui-user.sh and remove --skip-install, then restart it. --skip-install prevents the dependencies from being installed when adding new extensions.

modern quest Aug 4, 2023, 4:11 PM

#

brazen river Not in relation to the previous poster, i haven't tried adetailer, just wanted t...

Glad to hear it 👍

wind galleon Aug 7, 2023, 6:30 AM

#

I have gotten this to work before but today when i ran it I had #1137923208050393220 -- basically it threw an Easy Gui error, which -- well is the WHOLE entire reason you made the docker images XD -

I"m gonna try again either tonight or tomorow in case it was the machine i chose.

chrome trout Aug 7, 2023, 11:11 PM

#

what's the default password/token to open jupyterlab?

modern quest Aug 7, 2023, 11:12 PM

#

chrome trout what's the default password/token to open jupyterlab?

Its in the README

chrome trout Aug 7, 2023, 11:15 PM

#

ah, didn't see a readme before! thank you!

#Stable Diffusion UI with Dreambooth, ControlNet, Deforum, roop extensions preinstalled & Kohya_ss