Get Started

Installation

Clone the library source code:

git clone https://github.com/teocala/pihnn.git

and move to the library main folder.

Note

Before the installation, it is recommended to create an Anaconda environment:

conda env create -f environment.yml # the file 'environment.yml' already provides all dependencies
conda activate pihnn-env # 'pihnn-env' is the name of the newly created environment

Finally, run

pip install .

Warning

Make sure the name and version of the library are correctly detected during installation. There are currently some issues with some versions of pip and linux (e.g., here)

Run a test

The examples folder contains some tests that can be easily run from command line. For example,

python3 examples/simply_connected/laplace.py

will run the test on the Laplace equation and provide plots in the results folder.

Using CUDA

It is possible to exploit GPUs for accelerated processing if you have a CUDA-capable system with the CUDA toolkit installed. Furthermore, you need to install the CUDA enabled PyTorch by running:

conda install pytorch-cuda -c pytorch -c nvidia