Generate Your Images Faster by Installing Stable Diffusion With Google Colab!
ฝัง
- เผยแพร่เมื่อ 19 มิ.ย. 2024
- IMPORTANT NOTE: YOU CANNOT RUN THIS CODE WITHOUT UPGRADING TO COLAB PRO
How to install Stable Diffusion with Google Colab. Here we'll walk through installing Stable Diffusion from start to finish, including ControlNet Extensions.
📣📣📣I have just opened a Discord page to discuss SD and AI Art - common issues and news - join using the link: / discord
👀 Struggling to run Stable Diffusion locally? You can try DiffusionHub!
🤑Get 20% MORE DiffusionHub CREDIT using the link below or using the promo code LAURA20!
Link: diffusionhub.io?fpr=laura17
🤙🏻 Follow me on Medium to get my Newsletter:
- Get UNLIMITED access to all articles: / membership
- Laura: / lauracarnevali
- Intelligent Art: / intelligent
📌 Links:
Google Colab: colab.research.google.com/git...
GitHub: github.com/AUTOMATIC1111/stab...
Hugging Face: huggingface.co/runwayml/stabl...
00:00 Why install Stable Diffusion with Colab?
01:13 Open Google Colab
02:04 Connect Google Colab to Google Drive
03:25 Install/Update AUTOMATIC1111 repo
04:18 Install requirements
04:56 Model download/load
08:26 ControlNet extensions
10:50 Start Stable Diffusion
11:24 Final thoughts
#aiart #stablediffusion #generativeart #googlecolab #colab
Hi I am your new subscriber. first of all, thank you for creating these amazing tutorials. I managed to install SD on my m1 macbook air. I also followed your last tutorial to install SD on google Collab and got it working, however next day when i was trying to work on SD, it gave me an error. I want to ask if I need to follow the entire reinstallation process every time I wanted to work on SD through collabe or is there any way we can simply launch the SD without going through the entire installation process?
Hi Taruun, Thank you!! :) A downside of Colab is that you need to re-run the Colab notebook each time. This is because starting a new Colab session is similar to working on a fresh computer.
@@LaCarnevali thanks for answering my question. Now I need to make a choice between running SD on M1 vs. Collab. 😊
I would go with Colab :) quicker and less space consuming
@@LaCarnevali Hi! I have this issue too! May you explain what do I have to do to re-run the Colab notebook? If I restart a new colab page and I refollow all the steps in this video it says that I can't connect the GPU due to Colab limitations...
@@MeilinLi1 Hi Meilin, this is due to Colab GPU limitations. Unless you want to pay for Colab pro, you have to wait until the GPU recharges. Otherwise, you can run it without GPU, but it will be slower.
Thanks Laura, youv made my day.
The best tutorial about instal stable diffusion I find so far.
Simple but with a dev approach. with a slow and understable voice 👍
That was way easier than I was dreading, thanks a lot for the instructions!
I have tried this several times following other TH-cam videos and was unsuccessful. I just tried doing it following your steps. Your explanation is simple that even a 1st grader can understand and follow them. Thanks for the simple but informative video. You have just gained a subscriber 😍
this was so useful and straight to the point, thank you
after seeing few tutorial to install in google colab, this one really help me... thank you Laura..
Thank you so much, for the clarity in your tutorial/overviews. Good teaching, Molto Grazie Maestra Carnevali
This was extremely useful. Thank you. You made a lot more sense to me than all the reading I've done, so far. 😅
Grazie mille Laura for these tutorials! They are really clear and easy to follow, I've finally managed to use and understand SD and google colab on my mac thanks to you!🙏
Thanks so much for this tutorial! This was really helpful and clear! Liked and subscribed! One important nuance I've found out (maybe very obvious) is that to load Colab on a Mac is preferable in Chrome.
Wow, this is simple and straight to the point. You just earned a new subscriber. Can you also make a video tutorial about installing latest stable diffusion and its requirements locally on a windows computer? Thanks.
Thank you!! A weight off my back and promising content.
thank you so much! It finally works!
Thank you so much for this video.
I was literally planning to upgrade my setup and buy a new Nvidia graphic card (as I'm using an AMD one) just to be able to use Stable Diffusion locally.
Well, the upgrade has to wait. I'm going to have fun on Stable Diffusion, now ^_^
Thank you this was so helpful! Subscribed!
I was able to run everything without a sub to Colab Pro. Not sure if something changed, but I was able to receive a link and use the Web UI. Thanks for the great tutorial.
Thank you very much Laura for sharing this knowedge.
love all the help , amazing
Thanks for this useful video!
Hello Laura, thank you very much for your tutorial. It is very informative and easy to follow. I bought the 100 Google credits, to test it out a bit.
Really well explained. Thank you Laura.
These are so good thank you!
Very well done explanations! Thank you very much, hepled me a lot!!
Greetings and all the best!
Grazie mille, tutorial e metodo semplicissimo e funzionale. Complimenti, non vedo l'ora di aggiungere anche Deform. 👍
It contains deform
Nice video I didn't know about this option
thank you! thats super useful!!
Hello, a really good video, thanks for sharing your knowledge.
Truly has been one of the best methodologies to follow for Macs. The local installation worked too however, it was slow and cumbersome this one seems to be a bit better as it is on the cloud.
Beautiful look
Thank you so much Laura
I was using SD locally and I was thinking about buying a new gpu card, but after watching this video, I switched to Colab and bought extended Gdrive space. Good video, I'd love to see a video describing different I guess Colab notebooks because I've seen yours on Auto 1111 and I've seen one on Training ckpts, is there one for Koyhas? Or a video on traning Loras using Colab? I guess notebooks in this since are codes people have made to work using Colab.
Meriti minimo mezzo milione di iscritti. Bravissima.
Thanks for the video
You're awesome. Thank you.
Lovely videos!! I hope your channel goes off!!
You're a breath of fresh air. Those indian dudes mean well, but don't explain anything and produce sub par videos. You actually take us step by step. I know this video is old but it's super helpful. Liked and subscribed.
Thank you very much miss
I liked the method you used to explain each step for the installation of Stable Diffusion in Google Colab. If I wanted to add a new extension, what would be the method to incorporate it?
Thank you, Fernando :)
There should be a tab named "Extensions" when you start SD on Colab. From there, you can click on "Load from", and a list of extensions ready to be installed should come up.
You can watch this video if you have any doubts:
th-cam.com/video/SktvO_OtnOQ/w-d-xo.html
Thanks for sharing your experience with us. I have trained a SD + Dreambooth model and save a copy of the colab notebook at my Gdrive. Let's assume it is called model 1. If later I would like to generate a new model by training SD + Dreambooth + Lora + Latent Couple, how should I do to prevent the model 1from being overwritten or messed up with the new model, model 2 ? Any idea ? Thank you very much!
Thanks so much
Thank you for the video. I just was wondering.I understand when i use colab, I have to redownload it again. Do I have to download everything again or can i just download certain parts. Do i have to delete everything out when im done. One section was 28.1 gb I have 2tb storage but i dont want to just waste space if i dont need to. I also have Colab pro. Ty for any help you can give me
How do we reopen after the first installation, do we have to redownload the models or should we just copy the path of the model in our drive. And additionally can we add --listen code into command args?
Really cool is it possible To go deeper into control net and scribble?
I was thinking for kids perspective.
So kids can draw ✍️ anything and then the AI show their drawing into a real cool animation
Thanks so much. Really helpful. Do I have to follow those steps every time I restart my computer?
Yes, it should be quicker after the first time
One question. If I close the colab session and want to restar other day, I must repeat all the steps? The previous work done before will continue existing?
Yes, you have to restart the process again. This is because starting a new Colab session is similar to working on a fresh computer. The work you did before should be saved in your drive.
Hello Laura! My name is Lucas, and I'm sending you a message from Rio de Janeiro, in Brazil. I'm already a big fan of your amazing work with AI Art. I'm really impressed by the way. So, Iet me ask you a question... I already installed the "Model (21GB). I think for installing the v2_model, I should pay for the Colab Pro version, right?
Hi Lucas! Love Brazil, would love to pass by soon! :)
You may still be able to run them on the free version of Colab, as the virtual machine provides 100GB of disk space. However, keep in mind that the virtual machine also has memory and CPU limitations that may impact the performance of your models.
@@LaCarnevali Thanks a lot for your feedback! I don't know if you're coming for a business trip, but if you'll make a seminar or so, I would love to be there!
Amazing video, you explain things very well. I'm unable to complete the module step and need help, can I screen share with you on a live support session?
I think no
thanks, you saved me
gracias
Hey Laura, can you make a video of an alternative online platfrom to run comfyui on other than colab since I and a few others noticed that colab seems to have problems even with the paid version? Would be appreciated
Hi, thanks for the tutorial, Please how can I add a Deforum extension?
th-cam.com/video/SktvO_OtnOQ/w-d-xo.html Just published this! :)
Thanks you so much, i have a question! When i am connected for 2-3 hours in Google Colab, it says that i reached my limit for GPU... Does it resets after time?
Hi Jennie, it s not very clear how this works. You might get access to GPU again after a day, a few hours, or a few days. However, if you have another Google account, you might be able to run your notebook on there...
Thanks so much for this tutorial, there was a 504 Gateway Time-out once I clicked on the link, which step I missed or what should I do?
Yes google colab run few hours i think 9 hour before stop work and you may to wait a few hour or tomorrow to run it again.
Thanks for the video it is great. A question please: Do I have to redo all the steps each time I use it. Or has a link been saved somewhere that I click. Thanks
Bill G
yes, you need to go through all of the steps unfortunately
Ciao Laura, I have a silly question. How do you put a dark theme on sdxl on google colab?
hello Laura, can we use colab to speed the stable diffusion local?
Can these be used on mobile iPhone or iPad?
Can you make one with diffusers running entirely on Colab? I want it build one to make profile pictures given a personal photos, getting into this from yesterday, amazing!
You are awesome ❤❤
is it necessary to redo all these settings again and again to start sd ? de tells me that it does not find an interface
Yup
Really appreciate Laura sharing the step by step guide, quite helpful. I manage to complete all steps, however once I click the generated URL, it show the site but with the message 'no interface is running right now' . Have I missed a step?
Hi Adrew, reading online, it seems that sometimes it can happen and it's due to Gradio. Try to run the code by ticking the Use_localtunnel box. Let me know if it works :)
@@LaCarnevali thanks a lot for the advice, i could not find the box showing 'use_localtunnel box', instead I click 'use cloudfare' box and I was able to get to the web GUI, so the issue is resolved
@@fuzzygel Happy you sorted! :)
Another great video. I am curious about censorship. Do they censor what you can input with Collab, or is it like using it locally?
Exactly the same as using it locally
@Laura Carnevali thanks. Tried it out already. Having problems getting dreambooth installed on this
Hi, thank you for your tutorial. I followed the step-by-step to install SD on Collab and worked perfectly, but now I'm having trouble relaunching it. Do I need to run every single code again, (including Install/update Automatic 1111 repo, Requirements, Model download, etc.) or is there a different way to do it? When I try to run all codes again it gives me an error . I might be doing something wrong, so could you clarify the steps I need to take to relaunch? Thanks so much
Hi Bernat. Yes, you need to rerun all of the steps. I would recommend to close and reopen the notebook. Every time you open a new notebook on Colab you are connecting to a different computer. That's why you need to rerun all of the steps
@@LaCarnevali Hi, do you have to download all of the control net models each time too?
do i have to repeat all the same steps the next time i use it?
Hi Laura. This process was all this done on a Mac, right? Since it is a web based process could be done in any system, or not?
Exactly, you can use it on other systems too ;)
@@LaCarnevali Thanks! great work you are doing!
Good video! I’m wondering how to train my model on colab?
Given the time constraint (12h of continuous work), should use a small dataset.
Great tutorial - Diffusion installed and tested!!
Does anyone know of a good primer on using the GUI, though? I’m jumping in and messing around, but keep breaking things!
Thanks. I am working on building an app - however, this code is constantly updated, so I could spend more time updating the app each time instead of running the code. lol
If someone has another view please share :D
A deserved thumbs up from here :)
Hey Laura, thank you for the video. When I arrive at the Runtime type there is not the GPU option. Instead, I allows me to choose Python 3 or R. Can you tell me what to do?
Have you upgraded to Colab PRO?
I opened your google colab file in the description but why is the automatic1111 install part different from the one in this video?
Hi, the video was made a few months ago, so the notebook has changed since then
Hi Laura, excellent video, are you available for private (paid) consultations ?
Hello Laura, new subscriber to your amazi ng channel. I have a question about Mochi Diffusion, wouldn't that run faster on Mac apple silicon natively than stable diffusion web ui?
Yes it should - I haven't tried it though, so not sure if it has the same flexibility as Automatic1111. I will check that :)
when I press generate I see this "LayerNormKernelImpl" not implemented for 'Half' why ?
Thank you for your great tutorial!! Very easy to understand even for a non-english speaker like myself. I have a question though, my google drive only have 15GB but the ControlNet needs 21GB. Is there anyway to work around it?
Hi, thanks :) You can buy storage on Google Drive. Not sure if there is a free alternative.
Same question, is there any free option for installing the 21GB of ControlNet
@@Vtrescincodiez If you have a good PC, you can install it locally. I don't think you can get free external space for free.
You dont need al 21 gb of controlnet) You need to klick a scroll down menu button on the right of this 21gb package, and choose just one of its model (in my opinion, canny model is the best)/ Then you need to choose second model, and it should be also "canny". Both of them weights only about 1,5gb
how can I start the interface again ? radio says no interface is running but everything is installed in drive
Need to re-run the code...
Is it possible to install other models in the models directory with this tutorial?
You can download the models and upload them in the Drive before running the code.
You need to upload them under: sd > stable-diffusion-webui > models > Stable-Diffusion
I try to understand Colab. Is a Colab Notebook something like a project? Do i have redo all the installsteps and download the models each day i try to train stable diffusion? Is there no way to have a Alias and start the programm from the dock like PS or InDesign?
Yes, unfortunately you need to rerun the code every day, because every time you are running it on different computer. But you are not loosing the work previously done. Consider that the time saved on generating image is worth running the model again.
@@LaCarnevali Thank you for your help!
i like to use inpainting so do i need to install both those files or just one.
Can install only one. If you have both it's better, you can see different results using both.
Hello Laura Carnevali. Thank you for providing this tutorial. May I ask you something? Currently, Google Colab is a paid service and the cheapest option is $9.99, which includes 100 compute units. Do you know how many images can be produced with 100 compute units? Or how the calculation works? Thank you in advance :)
Hi, not sure - it really depends many things so not easy to estimate. You can try ComfyUI, which is free, or have a look at DiffusionHub (under payment but more stable than Colab, which loose connection very easily). If you want to try DiffusionHub, you have 30minutes free and then you need to pay credits --> can get 10% off with my link diffusionhub.io/?fpr=laura17
Hi Laura. Every time i try to run SD, i get an error that a module is missing. It started with FastAPI, but it just doesnt stop after that. cheers
This is a common issue from yday!
It looks like Colab updated Python to 3.9. TheLastBen is updating the dependencies but it will take a bit of time I suppose.
Followed the tutorial step by step, have everything installed, but while I try to work with the web UI, or generate nay image, it shows connection error. and does not work, I can't even change any settings P.s. My drive has enough empty space. Please help!!!
Hi. Im new to SD why do you have to upgrade to Colab Pro? Is it worth it or is there better opinions?
Hi Evan, it depends on what system you have. If you have a Windows, I would recommend using your computer (local installation). If you have a Mac, I would recommend using colab/runpod. Some are happy with Colab, some are not. Runpod seems to be winning between the two. Otherwise, you go for the pay-option, I can also recommend you Diffusion Hub. You have free 30 minutes if you want to try it, after need to buy credits diffusionhub.firstpromoter.com/login
Hi you didn't show what link you use to get the image of stable diffusionon google colab
do I need to rerun every step when I wanna run SD? I have a notification tell me that I can't make changes to this notebook?
Yes, you need to rerun all of the steps :/
ControlNet is now about 21GB with all models and it hangs while trying to install them all I guess because google drive space is limited?
Hi, I try to install the stable diffusion the way you explain but i got a problem with my google drive storage. my storage limit is only 15 GB. can i directly connect colab with my hard drive or i have to purchase storage from google
Hello! While Google Colab cannot directly access your local hard drive due to the fact that Colab notebooks run on Google's servers in the cloud, you can use a workaround by creating a server on your local machine that Google Colab can access (using ngrok I think). But then you will also need to modify the code in the Google Colab notebook. So it can be complicated :)
after installation in google colab does not generate an image
RuntimeError: 'LayerNormKernelImpl' not implemented for 'Half'
what to do?
Can you please explain in detail, how do i use the model link option to use custom model of choice?? Thank you
Helloo, :)
If the model is from Hugging Face, for example this URL link:
huggingface.co/runwayml/stable-diffusion-v1-5/blob/main/v1-5-pruned-emaonly.ckpt
Note: the link has to direct you to the checkpoint, like the above!
You need to simply copy and paste the URL link in the "MODEL_LINK" under the "Model Load/Download" Section.
If the model is on CivitAI, like this one: civitai.com/models/4823/deliberate
You need to right-click on the "Download" button and choose "Copy Link Address".
You then need to paste the link copied in the "MODEL_LINK" under the "Model Load/Download" Section.
Let me know if you have any issues :)
@@LaCarnevali i followed the steps you mentioned for hugging face url but i get an error saying 'wromg link, check the link is valid'
Edit : although the CivitAi one you mentioned worked! I'm just having problems with huggingface, i was trying to get waifu diffusion model
@@gamerx2.079 Can you send me the link you are using for the waifu? Seems weird cause the one I shared for HF is working for me. Make sure you are using the latest colab notebook: colab.research.google.com/github/TheLastBen/fast-stable-diffusion/blob/main/fast_stable_diffusion_AUTOMATIC1111.ipynb#scrollTo=p4wj_txjP3TC
Hello and thank you for the video. I think I am making a mistake at the beginning. I leave the "Shared_Drive" section empty, but then I get an error like "Google Drive file path not found" and I cannot run it. What should I do?
What if you connect manually? i.e., in Colab, if you go on the left bar, there is a folder. When you click on it, a left side bar should open. On top of this, click on the third folder ("mount drive").
Amazing tutorial! However I am running into a bit of a problem. It always works fine the first time I start it, but if I restart my runtime, I always encounter this error on the controlnet step: "OSError: [Errno 107] Transport endpoint is not connected: '/content/gdrive/MyDrive/sd/stable-diffusion-webui/extensions/sd-webui-controlnet/models'" on the control net step. I have tried everything, but the only way that I got it to work again why is by completely deleting the folder and reinstalling. Any advice would be very helpful!
Pretty annoying indeed - it seems to be a recent issue with GDrive. The turnaround would be to delete the lora/controlnet folder, run the code, and restore the folders deleted from the bin. Better than reinstalling? Not sure
Hi Laura, how long does the whole process take? I've been stuck on the last step for a long time.
Hi saviru, it should be quick. Try to restart the runtime and run again. Shouldn't take more than 5/10 minutes to download everything the first time. Should be even quicker after.
@@LaCarnevali Worked
hi thanks for the tutorial! after using it for a while I got a pop up that says "connection errored out" and when i refreshed the page it says No interface is running right now. Do you have any idea how to fix this?
What is the error log in the colab? There should be the source of the issue
Same here
@@chariskimbeki8342 Hi Charis, if you go back to the code below the Start Stable Diffusion Section, what is the error coming out exactly? If you cannot see an error, try to stop the code running, and run it again, some error should come out.
@@LaCarnevali Same here. For me, I was working in SD for an hour, and it happened. I didn't close the tab (nor did I close the colab tab). However, my Google Chrome puts to sleep tabs which are not in use. So I think the colab tab went to sleep.
Here is the error when I started SD in colab again:
---------------------------------------------------------------------------
ModuleNotFoundError Traceback (most recent call last)
in ()
4 import sys
5 import fileinput
----> 6 from pyngrok import ngrok, conf
7
8 Use_Cloudflare_Tunnel = False #@param {type:"boolean"}
ModuleNotFoundError: No module named 'pyngrok'
---------------------------------------------------------------------------
NOTE: If your import is failing due to a missing package, you can
manually install dependencies using either !pip or !apt.
To view examples of installing some common dependencies, click the
"Open Examples" button below.
---------------------------------------------------------------------------
hi, thanks for the detailed informations. unfortunetely i stuck at the ControlNet step and got "NameError: name 'capture' is not defined". I tried a few times from the beginning but cant pass this step. i have 200gb subscription for my Gdrive so I know it is not a storage issue. Do you have any suggestions?
This is probably a silly question, but are you sure you run all of the cells in the notebook? Including the requirements cell.
@@LaCarnevali yes, and I'm getting the same problem and I can't find any solution, Laura can you please share your Colab code for the ControlNet installation please
got the same error here too
Hey Laura, I've been using this google colab for a day now and it says I've run out of compute units is this supposed to happen if I'm not subscribed to the google colab?
yes, you have a max of GPU you can use on Colab, after that, you'll need to subscribe.
If you don't subscribe, it should still run, but slower and you should get GPU again after a month.
HI, the video was easy to follow. thank you. But i am encountering error in running controlNet extensions. can you help ?
what shd I do?
What errors?
@@LaCarnevali i sorted out the errors and got it to work, but now there is runtime error. everytime i go into stable diffusion there is a disconnection happpening. and then i start all the process again. it happened almost 15 times. cant get it to work. Plz help!
I'm facing an issue regarding the connection. The colab disconnects only after few minutes of connecting and before I click on generate or in the middle of the process it disconnects showing "Runtime disconnected" and I'm unable to use it. It was working perfectly fine but from 2 days I'm facing this issue. Please let me know how to stay connected??
Hi, Colab is not fully reliable with connection, I understand (mainly if your internet connection is not stable). If the issue is your internet connection, you can install SD on your computer locally. Otherwise, you can swap to RunPod, or Paperspace and see how it goes. I think Paperspace has a free option, but I have never used it.
what happened to the notebook? I try to run it, and all it says at the last step is
ModuleNotFoundError: No module named 'pytorch_lightning'
I take it you have to pay for the gpu? The only option is for me who hasn't paid is a T4 gpu, cpu, tpu.
Hi, Highly appreciable video and made one's life simpler. But every time I run the code, sometime it opens the interface and sometimes it shows Runtime Error. Even after opening, it shows the same error and don't generate image.
that's an issue with Colab unfortunately. If you are willing to pay for a subscription, you can try Diffusion Hub/ think Diffusion
Great tutorials and very clear and concise as well as knowlegeable ! - however, I have followed your tutorial re installing via Colab on an M1 Mac (I have also followed your tutorial on installing on M1 and it works up to a point but a lot of the models have not been downloaded due to storage space on my MacBook Pro and therefore I can't access a lot of the features) and yet keep running into an issue when trying to run SD since it gives me the same error message each time to do with a missing component.
Initially there was a warning re exceeding storage on G-Drive but I fixed this by upgrading to a higher tier sub' but now the missing component error is the problem.
I tried running it in Chrome instead of Safari but the same issue is there so I think this is why I opted to run SD locally.
I tried deleting SD & all components from my G-Drive and re-installing it under Colab but this hasn't fixed the issue - any suggestions ?
This is the error message I'm currently getting : 6 from pyngrok import ngrok, conf
7
8 Use_Cloudflare_Tunnel = False #@param {type:"boolean"}
ModuleNotFoundError: No module named 'pyngrok'">
Hi, this seems to be common to who doesnt have colab pro. Colab is not a free option anymore for SD. If you upgrade it should work. :(
@@LaCarnevali - Hmm...I just signed up for Colab Pro and the same error messages and stalls when trying to install...
@@LaCarnevali - Got it working eventually - the UI seems to running from a Gradio link, which I didn't expect for some reason.