dvaldivia / pytorch-notebook

Docker image with Jupyter, Pytorch and CUDA GPUs supports.

Jupyter Notebook with Pytorch

This docker image supports with jupyter, pytorch and cuda.

Run the container

Start the container with only CPU support:

docker run --rm -it  \
           -p 8888:8888  \
           -e JUPYTER_TOKEN=passwd  \
           tverous/pytorch-notebook:latest

Start the container with GPUs support:

docker run --rm -it  \
           --gpus all  \
           -p 8888:8888  \
           -e JUPYTER_TOKEN=passwd  \
           tverous/pytorch-notebook:latest

Start the container with volumes:

docker run --rm -it  \
           --gpus all  \
           -p 8888:8888  \
           -e JUPYTER_TOKEN=passwd \
           -v /local_vol:/docker_vol  \
           tverous/pytorch-notebook:latest

Others (Experimental)

Build the image with non-root users

To start the container with a non-root user, you need to build a new image that includes the designated user.

git clone https://github.com/Tverous/pytorch-notebook.git
cd pytorch-notebook/

Build a new image that incorporates a non-administrator user within.

Warning

The NOPASSWD option is enabled for the sudo command in the create-user.dockerfile file, signifying that no password is necessary to execute the sudo command. Modify this setting if it is not desired.

docker build --no-cache \
             -f create-user.dockerfile \
             --build-arg MY_UID="$(id -u)" \
             --build-arg MY_GID="$(id -g)" \
             --build-arg USER=demo \
             --build-arg HOME=/home/demo \
             -t tverous/pytorch-notebook:user \
             .

Start the container with the image you just builded.

docker run --rm -it  \
           -p 8888:8888  \
           -e JUPYTER_TOKEN=passwd  \
           tverous/pytorch-notebook:user

Where the argument MY_UID is the user id for the created user, MY_GID is the group id for the created user, USER is the name of the created user, and HOME is the home directory for the created user.

Please check the file create-user.dockerfile for details

Build the image with HTTPS supports

Modify the parameters to OpenSSL and PEM_FILE_PATH defined in the file https.dockerfile for your needs

Build a new image with HTTPS supports.

docker build --no-cache \
             -f https.dockerfile \
             -t tverous/pytorch-notebook:https \
             .

Start the container with the image you just builded.

docker run --rm -it  \
           -p 8888:8888  \
           -e JUPYTER_TOKEN=passwd  \
           tverous/pytorch-notebook:https

Now you can access your host with HTTPS.

Build the image with jupyter lab extensions

Jupyter Lab supports extensions to enhance its functionality.

Check awesome-jupyter for a list of awesome JupyterLab extensions and resources.

git clone https://github.com/Tverous/pytorch-notebook.git
cd pytorch-notebook/

Build a new image with Jupyter Lab extensions installed.

docker build --no-cache \
             -f jupyter-lab-extension.dockerfile \
             -t tverous/pytorch-notebook:extension \
             .

Start the container with the image you just builded.

docker run --rm -it  \
           -p 8888:8888  \
           -e JUPYTER_TOKEN=passwd  \
           tverous/pytorch-notebook:extension

Update the file jupter-lab-extension.dockerfile for other extensions you would like to install.

Launch Jupyter Notebook

When you start a notebook server with token authentication enabled (default), a token is generated to use for authentication.

This token is logged to the terminal, so that you can copy/paste the URL into your browser:

If you did not specify the token before starting the container, make sure to copy/paste the token logged on the terminal

[I 11:59:16.597 NotebookApp] The Jupyter Notebook is running at:
http://localhost:8888/?token=c8de56fa4deed24899803e93c227592aef6538f93025fe01

Make sure to update the localhost of the url to your remote server IP, if you are running the container remotely

Detach the logged context in the tty

Press Ctrl + p and Ctrl + q to detach the tty.

References

docker run --rm \                       # remove the container when it exits
           -it \                        # pseudo-TTY
           -p 8888:8888 \               # port forwarding: <Host>:<Container>
           --gpus all \                 # support all gpus (docker > 19.03)
           -v /local_vol:/docker_vol \  # volume: mapping local folder to container
           -e JUPYTER_TOKEN=passwd \    # Jupyter password: passwd
           -d tverous/pytorch-notebook:latest