Rancher Desktop

Download Rancher Desktop from ​https://rancherdesktop.io and open it on your local desktop.

Dockerhub Account

Create a Dockerhub account on ​https://hub.docker.com. You will also want to create a repository to push the custom images you build. Call it 'deepxde'

Remember to log in to your dockerhub account from your desktop. Enter docker login in your command line, and enter your password. You will only have to do this once.

Building Custom Image

Specifically for the DeepXDE package, you will need to create a folder on your local desktop, deepxde. Within this folder, create a dockerfile with vim Dockerfile and enter the following:

FROM public.ecr.aws/j1r0q0g6/notebooks/notebook-servers/jupyter-tensorflow-cuda-full:v1.3.0-rc.1

USER root

RUN rm /etc/apt/sources.list.d/cuda.list
RUN rm /etc/apt/sources.list.d/nvidia-ml.list

RUN apt-get update && apt-get install -y --no-install-recommends imagemagick libopenmpi-dev \
 && apt-get clean \
 && rm -rf /var/lib/apt/lists/*

USER $NB_UID

# install - requirements.txt
COPY --chown=jovyan:users requirements.txt /tmp/requirements.txt
RUN python3 -m pip install -r /tmp/requirements.txt --quiet --no-cache-dir  -f \
 https://www.paddlepaddle.org.cn/whl/linux/mkl/avx/stable.html \
 && rm -f /tmp/requirements.txt

Within the same deepxde folder on your local desktop, create a new file called requirements.txt by entering the following in command line: vim requirements.txt. Here are some packages you may want to include:

ipython
numpy
scipy
tensorflow
scikit-learn
matplotlib
scikit-optimize
seaborn
tensorflow-probability
torch
torchvision
torchaudio
jax
flax
optax
paddlepaddle-gpu==2.4.2.post112
jupyter
jupyterlab
notebook
pandas
mpi4py
deepxde
mat73

You can add or remove the packages you need as required.

Now, to build the custom image (for Mac), enter the following in your command line: docker buildx build --platform linux/amd64 -t <username>/deepxde:<tagname> . You will need to use your dockerhub username and you can choose any tagname you want. For instance, your tagname could be v1 for version 1 of the image you are building.

Pushing to Dockerhub

After building the image, push to Dockerhub: docker push <username>/deepxde:<tagname>

Kubeflow

Go to the Notebooks tab and create a new server. Select custom image and enter <username>/deepxde:<tagname>. Choose the number of CPUs and GPUs (NVIDIA), also use your existing volume. Once this is done, you will want to launch the server!