Installation Guide

Prerequisites

• python_requires=”>=3.9, <=3.13”

• Supported Operation Systems: Linux

• [NVIDIA CUDA support](https://developer.nvidia.com/cuda-downloads), with a runtime version >= 11.4.2 to support asynchronize stream computation

Binaries

HyperGP is available on PyPI and can be simply installed with:

pip install HyperGP

From Source

If you are installing from source, you will need:

• Python 3.12 or later

• A compiler that fully supports C++11, such as gcc (gcc 8.5.0 or newer is required, on Linux)

• NVIDIA CUDA support:, with a runtime version >= 11.4.2

An example of environment setup in Linux is shown below:

• You should create a conda environment with some dependencies first:

conda env create -n HyperGP -f environment.yml
conda activate HyperGP

• If you want to build a wheel in local, just run the following command:

python ./setup.py sdist bdist_wheel

• Then, you can use the HyperGP through the pip install (replace the {str} to the actual string of the whl in /dist):

pip install HyperGP-{str}.whl

• Or you can directly directly run the HyperGP through the source code, with the following command:

cd HyperGP
make compile
cd ..

• create a conda environment and install the dependencies by running the following command:

conda env create -n HyperGP -f environment.yml
conda activate HyperGP

• Activate the c++ compiler:

conda env config vars set CC=x86_64-conda-linux-gnu-gcc
conda env config vars set CXX=x86_64-conda-linux-gnu-g++
conda deactivate

• Build HyperGP:

conda activate HyperGP
cd HyperGP
rm -rf ./build/*
make compile
cd ..