Help with copying models into automatic docker #248

astralmedia · 2022-12-01T20:27:44Z

astralmedia
Dec 1, 2022

Hello! Thanks for the amazing work!

I have been trying for a few hours, and read everything I could find online, but I have been unable to figure out how to actually copy the model to the docker image.

I tried putting the model next to the dockerfile and adding this:

COPY sd-v1-4.ckpt ${ROOT}/models/Stable-diffusion

Any insight would be amazing! I need to have this happen on the docker side of things so the model is there every time it starts

Answered by DevilaN

Dec 7, 2022

So you are the one from reddit? :)

If you are happy with how this repo works, then simply create clean instance. Let docker build image. Let it run and try different options in webui, as some of them are downloading additional dependencies on first execution.

Later, when you are happy with all stuff just create your own image from running container with docker commit. This will export your running container with everything in it as another image.
Now it is only a matter of new Dockerfile starting FROM yourfancyimagename and use COPY commands to copy models and other mounted directories into specific locations. And viola... big blob image that you can distribute it for friends / colleagues…

View full answer

AbdBarho · 2022-12-01T21:47:00Z

AbdBarho
Dec 1, 2022
Maintainer

@astralmedia what you have done is correct, however, I override the folders in the docker compose:

stable-diffusion-webui-docker/docker-compose.yml

Line 7 in 8df9d10

- &v1 ./data:/data

and then mount the files to the entrypoint:

stable-diffusion-webui-docker/services/AUTOMATIC1111/entrypoint.sh

Lines 22 to 38 in 8df9d10

    
           MOUNTS["${ROOT}/models/Stable-diffusion"]="/data/StableDiffusion" 
        
           MOUNTS["${ROOT}/models/VAE"]="/data/VAE" 
        
           MOUNTS["${ROOT}/models/Codeformer"]="/data/Codeformer" 
        
           MOUNTS["${ROOT}/models/GFPGAN"]="/data/GFPGAN" 
        
           MOUNTS["${ROOT}/models/ESRGAN"]="/data/ESRGAN" 
        
           MOUNTS["${ROOT}/models/BSRGAN"]="/data/BSRGAN" 
        
           MOUNTS["${ROOT}/models/RealESRGAN"]="/data/RealESRGAN" 
        
           MOUNTS["${ROOT}/models/SwinIR"]="/data/SwinIR" 
        
           MOUNTS["${ROOT}/models/ScuNET"]="/data/ScuNET" 
        
           MOUNTS["${ROOT}/models/LDSR"]="/data/LDSR" 
        
           MOUNTS["${ROOT}/models/hypernetworks"]="/data/Hypernetworks" 
        
           MOUNTS["${ROOT}/models/deepbooru"]="/data/Deepdanbooru" 
        
           MOUNTS["${ROOT}/embeddings"]="/data/embeddings" 
        
           MOUNTS["${ROOT}/config.json"]="/data/config/auto/config.json" 
        
           MOUNTS["${ROOT}/ui-config.json"]="/data/config/auto/ui-config.json" 
        
           MOUNTS["${ROOT}/extensions"]="/data/config/auto/extensions"

if you want to put all models in the container (note: it will be huge!!!), you can try the following:

please note that this method has a lot of limitations and you will lose a lot of flexibility.

Run the app at least once to make sure the files in the data folder get populated.

Copy the data folder into services/AUTOMATIC1111

Add the following line

RUN mv /docker/data /data

exactly here at line 84 after the copy :

stable-diffusion-webui-docker/services/AUTOMATIC1111/Dockerfile

Lines 83 to 85 in 8df9d10

    
           COPY . /docker 
        
           RUN <<EOF

and then run

cd service/AUTOMATIC1111
docker build -t standalone-auto .

Note: it will take a lot of time and will use a lot of memory, so be careful!

you can now run your image directly, there is a lot of config that has to be done manually, which is why we use docker compose, but the command could look like this:

docker run -it \
   -p 7860:7860 \
   -v $PWD/output:/output \
   -e CLI_ARGS="--allow-code --medvram --xformers --enable-insecure-extension-access" \ 
   standalone-auto

I have not tried any of this, but I think it should work.

2 replies

astralmedia Dec 7, 2022
Author

Hey I was trying to get this working to this spec you shared above, and when I hit the RUN mv /docker/data /data it says:

mv: cannot stat '/docker/data': No such file or directory

The data folder is populated, and everything else is running properly. Any ideas?

astralmedia Dec 7, 2022
Author

fyi that error happens during the build command.

DevilaN · 2022-12-02T06:35:52Z

DevilaN
Dec 2, 2022

Why would you like to copy model into image? This seems like a some misconception behind this decision.

1 reply

astralmedia Dec 2, 2022
Author

I actually want to have the automatic docker download the model & requirements...trying to get automatic fully dockerized for a specific use case here at my studio.

astralmedia · 2022-12-02T23:14:18Z

astralmedia
Dec 2, 2022
Author

Ok, I am seeing what you mean....Seeing as I want to dockerize the automatic repo, you think I could Frankenstein the two dockers together and just have the automatic dockerfile download and prep those files on build?

Thanks for the feedback btw! Seems like its working so far.

0 replies

astralmedia · 2022-12-07T18:31:22Z

astralmedia
Dec 7, 2022
Author

Hey guys, I am pretty close here! When I run the image with sh and do ls command, it shows the automatic folder contents, including the webui.bat and webui.sh files...

however, using this at the end of my dockerfile is not working:

CMD webui.sh

I set the workdir at the start, copy & install everything, then try to run the webui.sh or webui.bat and it says "not found" every time for the bat or sh file. I can run the python webui.py file, but it doesnt set everything up properly.

Thanks for any feedback!

0 replies

DevilaN · 2022-12-07T20:43:28Z

DevilaN
Dec 7, 2022

So you are the one from reddit? :)

If you are happy with how this repo works, then simply create clean instance. Let docker build image. Let it run and try different options in webui, as some of them are downloading additional dependencies on first execution.

Later, when you are happy with all stuff just create your own image from running container with docker commit. This will export your running container with everything in it as another image.
Now it is only a matter of new Dockerfile starting FROM yourfancyimagename and use COPY commands to copy models and other mounted directories into specific locations. And viola... big blob image that you can distribute it for friends / colleagues not worrying that somebody asks "how to download model and where to place it".

5 replies

astralmedia Dec 8, 2022
Author

So currently I have the yourfancyimagename image made properly, and I am building the COPY models stuff right now...

how do I inherit the entrypoint or CMD that the FROM image used?

astralmedia Dec 8, 2022
Author

So think I figured out how to move the models, and start everything off, but it seems the cuda drivers are not installed, I keep getting this:

Warning: caught exception 'Found no NVIDIA driver on your system. Please check that you have an NVIDIA GPU and installed a driver

DevilaN Dec 8, 2022

how do I inherit the entrypoint or CMD that the FROM image used?

ENTRYPOINT and CMD are inherited by default the same as in image you are doing FROM in Dockerfile.
You can check it with docker image inspect yourfinalimagename

Warning: caught exception 'Found no NVIDIA driver on your system. Please check that you have an NVIDIA GPU and installed a driver

Can you check with docker run --rm --gpus all nvidia/cuda nvidia-smi if it shows driver installed?
Also on some systems docker do not use gpu at all by default and you might need to install nvidia-container-toolkit package (or whatever your distro names it)

astralmedia Dec 8, 2022
Author

Running that command says:

Unable to find image 'nvidia/cuda:latest' locally

Tho that is just me running it on my dev machine, not like in the image shell

astralmedia Dec 8, 2022
Author

Hey! Sorry I am a newb! Thank you both for all the help! I was able to figure it out, and even tho its embarassing, I'll post it here for anyone else who runs into this:

Don't forget to add --gpus=all on your docker run command ;)

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Help with copying models into automatic docker #248

{{title}}

Replies: 5 comments 8 replies

{{title}}

{{title}}

{{title}}

{{title}}

{{title}}

{{title}}

{{title}}

{{title}}

{{title}}

{{title}}

{{title}}

{{editor}}'s edit

{{editor}}'s edit

{{title}}

{{title}}

Select a reply

Help with copying models into automatic docker #248

astralmedia Dec 1, 2022

Replies: 5 comments · 8 replies

AbdBarho Dec 1, 2022 Maintainer

astralmedia Dec 7, 2022 Author

astralmedia Dec 7, 2022 Author

DevilaN Dec 2, 2022

astralmedia Dec 2, 2022 Author

astralmedia Dec 2, 2022 Author

astralmedia Dec 7, 2022 Author

DevilaN Dec 7, 2022

astralmedia Dec 8, 2022 Author

astralmedia Dec 8, 2022 Author

DevilaN Dec 8, 2022

astralmedia Dec 8, 2022 Author

astralmedia Dec 8, 2022 Author

astralmedia
Dec 1, 2022

Replies: 5 comments 8 replies

AbdBarho
Dec 1, 2022
Maintainer

astralmedia Dec 7, 2022
Author

astralmedia Dec 7, 2022
Author

DevilaN
Dec 2, 2022

astralmedia Dec 2, 2022
Author

astralmedia
Dec 2, 2022
Author

astralmedia
Dec 7, 2022
Author

DevilaN
Dec 7, 2022

astralmedia Dec 8, 2022
Author

astralmedia Dec 8, 2022
Author

astralmedia Dec 8, 2022
Author

astralmedia Dec 8, 2022
Author