PDMD Version 3.0

PDMD Python package to perform automatic machine-learning (ML) model training, deployment, and assessment for ultrafast high-fidelity molecular dynamics simulations.

New features vs 2.0:

1. Support an arbitrarily sized chemical system
2. Significantly accelerated benchmarking codes through the pre-computed connectivity cutoff matrix

PDMD Version 2.0

PDMD Python package to perform automatic machine-learning (ML) model training, deployment, and assessment for ultrafast high-fidelity molecular dynamics simulations.

New features vs 1.0:

1. Support multiple-GPU training under the framework of Pytorch Lightning
2. Integrated with SLURM for efficient workload management on supercomputers

PDMD Version 1.0

PDMD Python package to perform automatic machine-learning (ML) model training, deployment, and assessment for ultrafast high-fidelity molecular dynamics simulations.

Features:

1. Support CPU and CUDA GPU.
2. Support ML models for the prediction of system energy and atomic forces based on chemically adaptive graph neural network (ChemGNN).
3. Support training a unified ML model for multiple chemical systems.