Installation and environment

aenet_env

From the repository root, run python -m pip install -e .. Check the actual pyproject.toml for dependencies. Use aenet_env-install to copy packaged workflow templates into the working directory; existing files are protected by default. Use aenet_env-install --force only when intentional replacement is required. Verify the command is available in the installed release.

Build and install ænet

git clone https://github.com/atomisticnet/aenet.git
cd aenet
cmake -S . -B build -DBUILD_AENET=ON
cmake --build build --target build_all

Example with Intel oneAPI Fortran and MKL (compiler path is site-specific):

cd ~/aenet
git clone https://github.com/atomisticnet/aenet.git build_aenet
cd build_aenet
cmake -S . -B build \
  -DCMAKE_Fortran_COMPILER=/opt/intel/oneapi/compiler/2023.2.1/linux/bin/ifx \
  -DUSE_MKL=ON -DBUILD_AENET=ON
cmake --build build --target build_all -j 8
cmake --install build --prefix /home/hpc/aenet

The installation places generate.x, train.x, and predict.x under /home/hpc/aenet/bin/, with tools such as fingerprint.x, neighbors.x, trnset_info.x, and trnset2ASCII.x under tools/. After installation, run aenet config --set-aenet-path /home/hpc/aenet and verify the executable paths configured by the Python frontend. An unqualified cmake --install build may install to a different prefix.

External requirements

VASP (separate license), PAW datasets, vasp_util, ænet Python bindings, and optionally PyTorch are required for the relevant stages. b03 needs a cluster scheduler and job script. The standalone ænet-PyTorch path needs a separate trainer and format converters. Fortran and PyTorch training paths are not interchangeable by assumption.

Build the bilingual documentation locally

python -m pip install -e .
python -m pip install sphinx myst-parser sphinx-rtd-theme linkify-it-py
bash docs/build.sh

Output is under docs/_build/html/en/ and docs/_build/html/ja/.