DeePMD-kit
DeePMD-kit is a package written in Python/C++, designed to minimize the effort required to build deep learning-based models of interatomic potential energy and force field and to perform molecular dynamics (MD). This brings new hopes to addressing the accuracy-versus-efficiency dilemma in molecular simulations. Applications of DeePMD-kit span from finite molecules to extended systems and from metallic systems to chemically bonded systems.
Input Section
Inference in CP2K is performed through the DEEPMD section. As an example, the relevant section for DeePMD-kit is:
&DEEPMD
ATOMS W
ATOMS_DEEPMD_TYPE 0
POT_FILE_NAME DeePMD/W.pb
&END DEEPMD
where the W.pb
refers to the DeePMD model that was deployed using DeePMD-kit. An example for the
full input file can be found and on the regtests, see
DeePMD_W.inp
Input details
The tag ATOMS expects a list of
elements/kinds and
ATOMS_DEEPMD_TYPE expects
a list of their index that is consistent with the type_map
in DeePMD-kit parameters. If this is
not done unphysical results will be obtained. Spotting such issues is quite straightforward as the
energy is significantly wrong.
Compiling CP2K with Libdeepmd_c
Running with DeePMD-kit requires compiling CP2K with the libdeepmd_c library. For the CP2K binaries,
please install the toolchain using the flag --with-deepmd
, which would download libdeepmd_c from
DeePMD-kit Github release and compile. GPU support is enabled when CUDA envrionment exists.
Further Resources
For additional references on Deep Potential and DeePMD-kit see:
DeepMD paper Wang2018, Zeng2023 and code https://github.com/deepmodeling/deepmd-kit