nlesc-nano / swan

Statistical models to predict new materials

Screening Workflows And Nanomaterials

🦢 Swan is a Python pacakge to create statistical models using machine learning to predict molecular properties. See Documentation.

🛠 Installation

• Download miniconda for python3: miniconda (also you can install the complete anaconda version).
• Install according to: installConda.
• Create a new virtual environment using the following commands:
  • conda create -n swan
• Activate the new virtual environment
  • conda activate swan

To exit the virtual environment type conda deactivate.

Dependencies installation

• Type in your terminal:

  conda activate swan

Using the conda environment the following packages should be installed:

• install RDKit and H5PY:
  • conda install -y -q -c conda-forge h5py rdkit
• install Pytorch according to this recipe
• install Pytorch_Geometric dependencies.
• install DGL using conda

Package installation

Finally install the package:

• Install swan using pip: - pip install git+https://github.com/nlesc-nano/swan.git

Now you are ready to use swan.

Notes:

• Once the libraries and the virtual environment are installed, you only need to type conda activate swan each time that you want to use the software.