Yuriyzabegaev / solver_selector

Automated solver selection for simulation of multiphysics processes in porous media

Installation

A Docker image with the full environment to reproduce the experiments, including PorePy and PETSc, is available on Zenodo: 10.5281/zenodo.10407057. After downloading the image, execute these commands (Docker should be installed):

docker load -i solver_selector.tar.gz
docker run -dit --name solver_selector -p 8888:8888 solver_selector:latest
docker exec -it solver_selector /bin/bash

Note that -p 8888:8888 is a port forwarding to make the Jupyter Notebooks server inside the Docker container accessible from your browser to visualize the results of the experiments.

Alternative installation

It is recommended to use the Docker image. However, this is a recipe of manual installation.

1. Install PETSc with Hypre and PETSc4Py.
2. Install PorePy from the develop branch.
3. Clone this repository:

git clone https://github.com/Yuriyzabegaev/solver_selector.git
cd solver_selector

4. Install the dependencies:

pip install -r requirements.txt
pip install -r examples/requirements.txt

5. Make the necessary directories observable by Python (<SOLVER_SELECTOR_DIRECTORY> and <POREPY_DIRECTORY> are the directories of the solver selector and PorePy, respectively):

export PYTHONPATH=$PYTHONPATH:<SOLVER_SELECTOR_DIRECTORY>/src
export PYTHONPATH=$PYTHONPATH:<SOLVER_SELECTOR_DIRECTORY>/examples
export PYTHONPATH=$PYTHONPATH:<POREPY_DIRECTORY>/examples

6. Run pytest to verify the installation:

pytest <SOLVER_SELECTOR_DIRECTORY>

Reproducing the experiments

The code to reproduce the experiments is in the examples directory. To run them all, use:

cd examples
bash run.bash <NUM_REPEATS>

where <NUM_REPEATS> is an integer representing how many times each experiment should be repeated. Alternatively, use run.bash scripts at the subdirectories of the experiments. The results are visualized in the results_*.ipynb notebooks. The Jupyter Notebooks server is one of the dependencies of the experiments, so it should be accessible by the command:

jupyter notebook --allow-root --ip 0.0.0.0 --no-browser

This command will provide a link to the server, such as: http://127.0.0.1:8888/tree?token=XXXX. Click on it, and it should open the Jupyter Notebook in the browser.

TODO:

• Reference to the paper
• Citing