PDEnsorflow is a library developed under TensorFlow 2.X to solve Partial dfferential equations.
Since version 1.2, it implements finite differences and finite element solvers.
Install TensorFlow 2.X using anaconda. First create an environment and activate it; e.g.:
conda create --name PDEnsorflow python=3.9
conda activate PDEnsorflow
If you want to just install TensorFlow, follows this link.
To install PDEnsorflow, proceed as follows:
Install cudatoolkit version 11.2 and cudnn version 8.1.0, as follows:
conda install -c conda-forge cudatoolkit=11.2 cudnn=8.1.0
Set up the LD_LIBRARY_PATH to the conda environment:
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$CONDA_PREFIX/lib/
Finally, cd to PDEnsorflow main directory and install with pip:
cd PDEnsorflow
python -m pip install -e .
Old version: Once the environment is activated, source the init.sh file in the main directory of PDEnsorflow:
source init.sh
New Version: The paths are authomatically set within __init__.py.
Then, launch one of the examples; e.g.:
cd Tests/Fenton_atria
python fenton.py
Note: *This run PDEnsorflow under GPU, provided that libraries are correctly installed. Otherwise, it will run under standard CPU. In the examples, the console will show under wich device the code is executed.