NVIDIA explains how to unlock Tensor Core performance for machine learning interatomic potentials
AINVIDIA's Technical Blog describes how to speed up machine learning interatomic potentials using Tensor Cores. The post frames the work against quantum-mechanical methods such as density functional theory, which give accurate energies and forces but are costly for large systems and long simulations.
Source details
NVIDIA Technical Blog · Anoushka Bhutani · developer.nvidia.comPublished · added here
Original title: Unlocking Tensor Core Performance for Machine Learning Interatomic Potentials