This code interfaces with Uni-Mol, DPA3, and some other neural networks (to be added), and learns properties, such as functional parameters and Hubbard-U values, from geometries.
Build the package with the following steps:
- create virtual environment
- install dependencies in the order DeePMD-kit (torch backend), then Huggingface, UniMol-Tools, finally this package.
# create virtual environment
# such as conda
conda create -n mlxc python=3.10
# then activate it
conda activate mlxc
# install dependencies
# export CXX11_ABI_FLAG=1 (consistent with DeePMD-kit)
export CXX11_ABI_FLAG=1
# install PyTorch cpu version
pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cpu
# install DeePMD-kit
pip install git+https://github.com/deepmodeling/[email protected]
# install DPData
pip install dpdata
# install Huggingface
pip install huggingface_hub
# install UniMol-Tools
pip install unimol_tools --upgrade
# install this package
pip install .# run unit test
bash .github/scripts/ut_driver.sh