Run in Google Colab

[monorural]

Hi there, I just came across this repo and am very excited to try it out. Is there a Google Colab available for running it?
If not, it would be awesome to have a notebook up to date in Google Colab, so that it can be run in the cloud.

This is an awesome example, very useful for me and lots of people who don't own strong enough GPUs to run it local:

https://github.com/camenduru/stable-diffusion-webui-colab

Thanks, this looks very promising! 🙌

[vladmandic]

thanks, but i haven't made a collab notebook or tested if its compatible - just not enough time.

[econundrum]

If I can find sometime this weekend, I might have a try and see if I can get it working.

I'm no expert on colab either but I've used it a little bit, I have a card good enough to produce images but not for training so having dreambooth built in along with other training methods is an advantage.

[monorural]

Thank you very much, I'm very interested to trying it out. Colab is an awesome tool, I've been using it since Disco Diffusion came out last year, and the amount of images I could create were huge thanks to it. Very reliable and having alternate Google accounts does the trick if you run out of computing power given for free.

[econundrum]

Colab only seems to support Python 3.9 at the moment so I'm dead in the water, though there may be hacks to get around this.

[vladmandic]

i'll relax requirements to allow python 3.9, but i cannot test it. check for update on github in ~1h.

[econundrum]

Sounds like they are upgrading colab to 3,10.x in the next couple of weeks anyway so might not be worth it.

[econundrum]

Got it running on colab, just want to tidy up a bit.

[monorural]

Edit only affects free users.

I'm referring to this:

https://www.reddit.com/r/StableDiffusion/comments/12tj22r/official_statement_from_the_person_in_charge_of/
https://www.reddit.com/r/StableDiffusion/comments/12trnw9/now_that_they_started_banning_stable_diffusion_on/

Google is starting to ban the usage of WebUIs with Google Colab, aiming mostly at stable-diffusion-webui because the usage just keeps growing and got huge

[AI-Casanova]

*free usage

[monorural]

you are right, my mistake

[cenahum]

They will not ban the users, but they can shut down your app in a minute or 2, maybe you can run for a while, this only happens because people do not respect the free thing, there is a lot of this kind of cases, they abuse and everybody pay, Midjourney is another example, it's a shame, but is reality. Anyway, people will now move to Kaggle and Gradiente and soon this will do the same. Anyway, there are some cheap options that we can use, or we can install in a desktop or in a notebook.

[cenahum]

Ok, I see that they change the mind, and they can Ban user if try to use SD webui, but I read that they will let use SD with Gradio.

[southbayjay]

Hey folks! Is there any update on the Colab notebook for this?

[spreck]

I found this code on some reddit comment the other day and it seemed to work fine:

#@title Connect to Google Drive
from google.colab import drive
drive.mount('/content/drive')

#@title V2 - Clone Vladmandic SD
%cd /content/
!git clone https://github.com/vladmandic/automatic

#@title Extensions install (optional)
%cd /content/automatic/extensions/
!git clone https://github.com/fkunn1326/openpose-editor
!git clone https://github.com/mcmonkeyprojects/sd-dynamic-thresholding
!git clone https://github.com/kohya-ss/sd-webui-additional-networks.git
!git clone https://github.com/Klace/stable-diffusion-webui-instruct-pix2pix
!git clone https://github.com/KohakuBlueleaf/a1111-sd-webui-locon
!git clone https://github.com/opparco/stable-diffusion-webui-composable-lora
!git clone https://github.com/opparco/stable-diffusion-webui-two-shot
!git clone https://github.com/nonnonstop/sd-webui-3d-open-pose-editor
!git clone https://github.com/richrobber2/canvas-zoom
!git clone https://github.com/hako-mikan/sd-webui-regional-prompter
!git clone https://github.com/ilian6806/stable-diffusion-webui-state
!git clone https://github.com/jtydhr88/sd-canvas-editor

#@title Controlnet 1.1 models install (optional, need huggingface token)
user_token = "" #@param {type:"string"}
user_header = f""Authorization: Bearer {user_token}""
%cd /content/
!mkdir -p ControlNet
!wget --no-clobber --header={user_header} https://huggingface.co/lllyasviel/ControlNet-v1-1/resolve/main/control_v11e_sd15_ip2p.pth -O /content/ControlNet/control_v11e_sd15_ip2p.pth
!wget --no-clobber --header={user_header} https://huggingface.co/lllyasviel/ControlNet-v1-1/resolve/main/control_v11e_sd15_ip2p.yaml -O /content/ControlNet/control_v11e_sd15_ip2p.yaml
!wget --no-clobber --header={user_header} https://huggingface.co/lllyasviel/ControlNet-v1-1/resolve/main/control_v11e_sd15_shuffle.pth -O /content/ControlNet/control_v11e_sd15_shuffle.pth
!wget --no-clobber --header={user_header} https://huggingface.co/lllyasviel/ControlNet-v1-1/resolve/main/control_v11e_sd15_shuffle.yaml -O /content/ControlNet/control_v11e_sd15_shuffle.yaml
!wget --no-clobber --header={user_header} https://huggingface.co/lllyasviel/ControlNet-v1-1/resolve/main/control_v11f1p_sd15_depth.pth -O /content/ControlNet/control_v11f1p_sd15_depth.pth
!wget --no-clobber --header={user_header} https://huggingface.co/lllyasviel/ControlNet-v1-1/resolve/main/control_v11f1p_sd15_depth.yaml -O /content/ControlNet/control_v11f1p_sd15_depth.yaml
!wget --no-clobber --header={user_header} https://huggingface.co/lllyasviel/ControlNet-v1-1/resolve/main/control_v11p_sd15_canny.pth -O /content/ControlNet/control_v11p_sd15_canny.pth
!wget --no-clobber --header={user_header} https://huggingface.co/lllyasviel/ControlNet-v1-1/resolve/main/control_v11p_sd15_canny.yaml -O /content/ControlNet/control_v11p_sd15_canny.yaml
!wget --no-clobber --header={user_header} https://huggingface.co/lllyasviel/ControlNet-v1-1/resolve/main/control_v11p_sd15_inpaint.pth -O /content/ControlNet/control_v11p_sd15_inpaint.pth
!wget --no-clobber --header={user_header} https://huggingface.co/lllyasviel/ControlNet-v1-1/resolve/main/control_v11p_sd15_inpaint.yaml -O /content/ControlNet/control_v11p_sd15_inpaint.yaml
!wget --no-clobber --header={user_header} https://huggingface.co/lllyasviel/ControlNet-v1-1/resolve/main/control_v11p_sd15_lineart.pth -O /content/ControlNet/control_v11p_sd15_lineart.pth
!wget --no-clobber --header={user_header} https://huggingface.co/lllyasviel/ControlNet-v1-1/resolve/main/control_v11p_sd15_lineart.yaml -O /content/ControlNet/control_v11p_sd15_lineart.yaml
!wget --no-clobber --header={user_header} https://huggingface.co/lllyasviel/ControlNet-v1-1/resolve/main/control_v11p_sd15_mlsd.pth -O /content/ControlNet/control_v11p_sd15_mlsd.pth
!wget --no-clobber --header={user_header} https://huggingface.co/lllyasviel/ControlNet-v1-1/resolve/main/control_v11p_sd15_mlsd.yaml -O /content/ControlNet/control_v11p_sd15_mlsd.yaml
!wget --no-clobber --header={user_header} https://huggingface.co/lllyasviel/ControlNet-v1-1/resolve/main/control_v11p_sd15_normalbae.pth -O /content/ControlNet/control_v11p_sd15_normalbae.pth
!wget --no-clobber --header={user_header} https://huggingface.co/lllyasviel/ControlNet-v1-1/resolve/main/control_v11p_sd15_normalbae.yaml -O /content/ControlNet/control_v11p_sd15_normalbae.yaml
!wget --no-clobber --header={user_header} https://huggingface.co/lllyasviel/ControlNet-v1-1/resolve/main/control_v11p_sd15_openpose.pth -O /content/ControlNet/control_v11p_sd15_openpose.pth
!wget --no-clobber --header={user_header} https://huggingface.co/lllyasviel/ControlNet-v1-1/resolve/main/control_v11p_sd15_openpose.yaml -O /content/ControlNet/control_v11p_sd15_openpose.yaml
!wget --no-clobber --header={user_header} https://huggingface.co/lllyasviel/ControlNet-v1-1/resolve/main/control_v11p_sd15_scribble.pth -O /content/ControlNet/control_v11p_sd15_scribble.pth
!wget --no-clobber --header={user_header} https://huggingface.co/lllyasviel/ControlNet-v1-1/resolve/main/control_v11p_sd15_scribble.yaml -O /content/ControlNet/control_v11p_sd15_scribble.yaml
!wget --no-clobber --header={user_header} https://huggingface.co/lllyasviel/ControlNet-v1-1/resolve/main/control_v11p_sd15_seg.pth -O /content/ControlNet/control_v11p_sd15_seg.pth
!wget --no-clobber --header={user_header} https://huggingface.co/lllyasviel/ControlNet-v1-1/resolve/main/control_v11p_sd15_seg.yaml -O /content/ControlNet/control_v11p_sd15_seg.yaml
!wget --no-clobber --header={user_header} https://huggingface.co/lllyasviel/ControlNet-v1-1/resolve/main/control_v11p_sd15_softedge.pth -O /content/ControlNet/control_v11p_sd15_softedge.pth
!wget --no-clobber --header={user_header} https://huggingface.co/lllyasviel/ControlNet-v1-1/resolve/main/control_v11p_sd15_softedge.yaml -O /content/ControlNet/control_v11p_sd15_softedge.yaml
!wget --no-clobber --header={user_header} https://huggingface.co/lllyasviel/ControlNet-v1-1/resolve/main/control_v11p_sd15s2_lineart_anime.pth -O /content/ControlNet/control_v11p_sd15s2_lineart_anime.pth
!wget --no-clobber --header={user_header} https://huggingface.co/lllyasviel/ControlNet-v1-1/resolve/main/control_v11p_sd15s2_lineart_anime.yaml -O /content/ControlNet/control_v11p_sd15s2_lineart_anime.yaml
!wget --no-clobber --header={user_header} https://huggingface.co/lllyasviel/ControlNet-v1-1/resolve/main/control_v11u_sd15_tile.pth -O /content/ControlNet/control_v11u_sd15_tile.pt
!wget --no-clobber --header={user_header} https://huggingface.co/lllyasviel/ControlNet-v1-1/resolve/main/control_v11u_sd15_tile.yaml -O /content/ControlNet/control_v11u_sd15_tile.yaml

#@title Launch Vladmandic webui (GPU mode)
%cd /content/automatic/
!COMMANDLINE_ARGS="--no-hashing --share --noupdate --skip-extensions --gradio-auth login:pass --ckpt /content/drive/MyDrive/SDmodels/clarity_19-pruned.safetensors --ui-config-file /content/drive/MyDrive/1my/vlad-ui-config-my.json --ui-settings-file /content/drive/MyDrive/1my/vlad-config-my.json" REQS_FILE="requirements.txt" python launch.py

[mykeehu]

Yes, it's based on my colab, since then I removed the --noupdate --skip-extensions options, because need the extension update.

But what I noticed is problematic on Colab:

• I can't change the model, it drops the Gradio connection
• I can't load SD 1.5 inpainting model, because as soon as I upload an image, the colab stops working.

But the image generation works, and I can change the model by changing the path in --ckpt.

[RimorCosmicam]

Hey there!

I've been using it for a while with Colab but didn't see the request, I have a single-cell colab to run/update it, however, you'll have to self-host the models on GDrive (I sync it with my computer, so it's simpler for me to add models and Loras), just run it, add your models to their respective folders, and you're good to go.

Vladmandic A1111

[cenahum]

Hi @RimorCosmicam
Thank you very much for sharing

[BinaryQuantumSoul]

Hey there!

I've been using it for a while with Colab but didn't see the request, I have a single-cell colab to run/update it, however, you'll have to self-host the models on GDrive (I sync it with my computer, so it's simpler for me to add models and Loras), just run it, add your models to their respective folders, and you're good to go.

Vladmandic A1111

Sadly it doesn't work. I have this error:

ImportError: cannot import name 'get_device' from 'basicsr.utils.misc'
(/usr/local/lib/python3.10/dist-packages/basicsr/utils/misc.py)

[Protonk]

This will run on the very first update, but subsequent runs get stuck after listing the Python version. I have no idea why this happens, no error messages are emitted, it just stalls before the 02:58:53-632197 INFO Version: 22e58809 Sat May 6 08:51:23 2023 -0400 string.

[AI-Casanova]

@Protonk just figured out that that is an issue with git since they upgraded colab to python 3.10

Add --skip-git and you'll be golden.

[Protonk]

AI-Casanova This works to solve that problem, though after a few server restarts the process fails as it cannot find clip.

[wa2el]

first of all thanks very much as everything is almost working perfectly

my problem is that CodeFormers isn't working

16:56:00-752061 INFO     Installing repositories                                
16:56:47-138300 ERROR    Error running git: None / -C                           
                         "/content/drive/MyDrive/A1113/repositories/CodeFormer" 
                         checkout c5b4593074ba6214284d6acd5f1719b6c5d739af      
16:56:47-140723 ERROR    Local changes detected: check setup.log for details

any idea how to solve this?

[cenahum]

Hi, I'm also facing some errors: Error running git: None / -C
"/content/drive/MyDrive/A1111/repositories/CodeFormer"
checkout
ERROR Local changes detected: check setup.log for details

ImportError: cannot import name 'get_device' from 'basicsr.utils.misc'
(/usr/local/lib/python3.10/dist-packages/basicsr/utils/misc.py)

[BinaryQuantumSoul]

Same issue here

[AI-Casanova]

CodeFormer has a couple annoying issues, but they are solvable.

/usr/local/lib/python3.10/dist-packages/basicsr/utils/misc.py needs to be modified, as the CodeFormer people modified the package away from the one you can pip install.

[AI-Casanova]

These are the two fixes needed

#81 (reply in thread)

[vladmandic]

you may want to check out https://colab.research.google.com/drive/126cDNwHfifCyUpCCQF9IHpEdiXRfHrLN
and report issues directly at https://github.com/Linaqruf/sd-notebook-collection - in general, those are best google colab notebooks and he offered to add one for this repo!

[BinaryQuantumSoul]

Nice, keep up with collabs :)

[BinaryQuantumSoul]

you may want to check out https://colab.research.google.com/drive/126cDNwHfifCyUpCCQF9IHpEdiXRfHrLN and report issues directly at https://github.com/Linaqruf/sd-notebook-collection - in general, those are best google colab notebooks and he offered to add one for this repo!

Shouldn't you also ask on https://github.com/TheLastBen/fast-stable-diffusion ?
He's waiting for the repo to "be well established" before adding it

[vladmandic]

author says its on the way: TheLastBen/fast-stable-diffusion#2062 (comment)

[wordbrew]

Honestly, I'm not hopeful for TheLastBen properly incorporating vladmandic. I have used Fast A1111 on colab for a few months now and it actually boots and runs slower than vladmandic on colab. I can't imagine TheLastBen's customizations to A1111 will improve vladmandic more than anything you've already done.

I've got a decent colab notebook built out that follows much of the same build out that TheLastBen has for his colab. Just need to find the time to incorporate the Controlnet models and I'm good.

[eucdee]

you may want to check out https://colab.research.google.com/drive/126cDNwHfifCyUpCCQF9IHpEdiXRfHrLN and report issues directly at https://github.com/Linaqruf/sd-notebook-collection - in general, those are best google colab notebooks and he offered to add one for this repo!

this colab is great out of the box, ive tried it yesterday with very simple promts and got stunning results 🗡️
so ive installed the github version localy and sadly its way different with many options still disabled.
sure i dont even know what 97% of these options really do yet but it would be nice to have them ;)
the authors notebook Cagliostro Colab UI is amazing aswell... i believe it uses anapnoes fork
i havent had time to check anapnoes fork but it would be nice to know the difference between the two.
ive tried setting up my local install to work like the colabs but i guess it would be better to try and run the notebook localy.
thats what will do now...

[wordbrew]

If anyone wants to use a colab, here is the one I built that Frankensteined some of TheLastBen's colab cells. It has some errors that I haven't bothered to fix yet (like the one @AI-Casanova solved for Codeformers). But it should work.

A few things to note:

1. This is for 1.5 SD. I took out anything pertaining to 2.x
2. I have not tested the ControlNet cell simply because I already have the 20gig files installed elsewhere. It should work though...

https://github.com/wordbrew/vladcolab/blob/main/Vladmandic_A1111.ipynb

I'm new(er) to Git, so sorry if this isn't right way to share the colab. I had built a gist, but it doesn't appear to allow you to easily copy the colab to your gdrive. This git repo does. I probably won't be maintaining it though.

[djwashout]

Humm. I'm new to all of this but let me see if i understand. From what i have gathered a notebook is just a notebook that runs commands in colab, (like on old bat file) and a repository is the actual code online in github that can install or execute? Sorry last time i had to learn tech was dos 5.0 running win.exe lol. its been a while! ...but if im right above. then damn. I just need to google the hell out of how to write a google colab notebook and im set. I was under the impression vladmandic had to be installed on a local machine...

[wordbrew]

Just view the thread. There are multiple colab notebooks already shared.

But yes, generally speaking, you're right.

And don't bother googling. Use ChatGPT to help you code the notebook if you want.

[southbayjay]

You can install it on a local machine, it was intended for that purpose originally but some of us don’t have the compute power to run it locally so we use other sources like Google Colab. If you have a capable pc to run it then jump on YouTube and look up Olivio tutorials. He did a pretty straightforward tutorial on installing it locally.

…

On Tue, May 2, 2023 at 11:00 PM djwashout ***@***.***> wrote: Humm. I'm new to all of this but let me see if i understand. From what i have gathered a notebook is just a notebook that runs commands in colab, (like on old bat file) and a repository is the actual code online in github that can install or execute? Sorry last time i had to learn tech was dos 5.0 running win.exe lol. its been a while! ...but if im right above. then damn. I just need to google the hell out of how to write a google colab notebook and im set. I was under the impression vladmandic had to be installed on a local machine... — Reply to this email directly, view it on GitHub <#81 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/A7JVNUHAGA6XESADHE6YHVTXEHYAXANCNFSM6AAAAAAWYS75CE> . You are receiving this because you commented.Message ID: ***@***.***>