DeePMD-kit

DeePMD-kit is a package written in Python/C++, designed to minimize the effort required to build deep learning based model 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.

The NGC DeePMD-kit container builds upon the NGC TensorFlow container, adding DeePMD-kit with LAMMPS plugin support.

System Requirements

• Pascal(sm60), Volta(sm70), or Ampere(sm80) NVIDIA GPU(s)
• x86_64 CPU supporting avx2_256 instruction set
• CUDA driver version >= 510.39.01, -or- r418(>=40.04), r450(>=36.06), r460(>=27.04), r470(>=57.02)

Running with Docker

Start an interactive session in the container

docker run -it --rm --gpus all --shm-size=1g --ulimit memlock=-1 nvcr.io/hpc/deepmd-kit:2.1.1

Fetch the DeePMD-kit example data

git clone https://github.com/deepmodeling/deepmd-kit.git

Train the example

cd deepmd-kit/examples/water/se_e2_a
dp train input.json

Freeze the model

dp freeze -o frozen_model.pb

Run the LAMMPS example simulation

cd ../lmp
cp ../se_e2_a/frozen_model.pb .
mpirun --allow-run-as-root -n 1 lmp -k on g 1 -sf kk -pk kokkos cuda/aware on neigh full comm device -in in.plugin.lammps

Suggested Reading

• DeePMD-kit documentation
• Deep Learning Frameworks Documentation - Running A Container
• NGC Container User Guide