Installation

Environment

This section describes how to prepare the environment necessary for running FS_GPlib.

Python Version

FS_GPlib is developed and tested with:

• Python 3.10

We recommend using a virtual environment to avoid conflicts with existing packages.

Create and activate a virtual environment:

conda create -n fs_gplib_env python=3.10
conda activate fs_gplib_env

Install

Install from TestPyPI

pip install -i https://test.pypi.org/simple/ --extra-index-url https://pypi.org/simple fs-gplib

Required Packages

If you encounter environment errors, try installing (or pinning) the following dependencies first, and then install fs-gplib:

• python==3.10

• torch==2.1.2

• torch_geometric==2.5.3

• numpy==1.24.1

• tqdm==4.64.1

• scipy==1.10.0

FS_GPlib requires PyTorch and PyTorch Geometric (PyG). torch_scatter is additionally required when you use models that depend on scatter operators (currently HKModel and WHKModel in fs_gplib.Opinions). You can install PyTorch/PyG by following the official instructions:

• PyTorch: https://pytorch.org/get-started/locally/

• PyG: https://pytorch-geometric.readthedocs.io/en/2.5.3/install/installation.html

If you need torch_scatter, install the wheel that matches your PyTorch and CUDA/CPU build. For compatibility details and wheel links, see:

• https://data.pyg.org/whl/

Example (CPU; change version tag to match your environment):

pip install torch-scatter -f https://data.pyg.org/whl/torch-2.1.2+cpu.html

Test Installation

To verify the installation was successful, try importing the library in Python:

import fs_gplib
print(fs_gplib.__version__)

If no errors are raised, the installation is complete.