Totten

Build Docker image from Totten

FROM tensorflow/tensorflow:latest-gpu

# install - requirements.txt
COPY --chown=jovyan:users requirements.txt /tmp/requirements.txt
RUN python3 -m pip install -r /tmp/requirements.txt --no-cache-dir  -f \
 <https://www.paddlepaddle.org.cn/whl/linux/mkl/avx/stable.html> \
 && rm -f /tmp/requirements.txt
RUN python3 -m pip install paddlepaddle-gpu==3.2.1 -i <https://www.paddlepaddle.org.cn/packages/stable/cu118/>
RUN apt-get update && \
    apt-get install -y git
RUN git clone <https://github.com/lululxvi/deepxde.git> && cd deepxde && python3 -m pip install -e .

This image is build base on TensorFlow gpu Docker.

ipython
numpy
scipy
tensorflow>=2.11.0
scikit-learn==1.4.2
matplotlib
tensorflow-probability[tf]>=0.19.0     # to use tensorflow 2
torch
torchvision
torchaudio
jax
flax
optax
paddlepaddle-gpu
pandas
deepxde
mat73
netCDF4
h5py

To build, go to the folder containing the two files above, run

docker buildx build --no-cache --platform linux/amd64 -t chenggongdartmouth/pinniclev1.0.3 .

Run Docker on Totten

Try to run the docker image on Totten first:

docker run --rm -it --entrypoint bash chenggongdartmouth/pinnicle:v1.0

and in the docker, check the version of DeepXDE, TensorFlow, etc. You can use pip list to show all the packages, or use something like:

python3 -c "import deepxde as dde;print(dde.__version__)"

to show one specific package.

Push to Docker Hub

After building, push the image to Docker Hub:

docker image push chenggongdartmouth/pinnicle_ls:v2.0

Lonestar6

After pushing your Docker image on to Docker Hub, you can download it to TACC’s Lonestar6: https://docs.tacc.utexas.edu/hpc/lonestar6/

Build image on Lonestar6

You will need to request a gpu node on Lonestar6, the best way to build image is to use interactive mode: