The benchmark of LiTEN
Description
This directory contains the raw and processed data for benchmarking the LiTEN model on training and testing tasks. Please note that in the rmd17 and md22 datasets, the energies stored in the .pt files are in kcal/mol, and the forces are in kcal/(mol·Å). However, during training and testing, the printed loss values are converted to meV and meV/Å, respectively.
For the Chignolin dataset, the .pt files also use kcal/mol for energies and kcal/(mol·Å) for forces. During its training, the units are kept consistent and no conversion is applied.
Files
chignolin.zip
Additional details
References
- Christensen AS, von Lilienfeld OA. On the role of gradients for machine learning of molecular energies and forces. Machine Learning: Science and Technology 1, 045018 (2020).
- Chmiela S, et al. Accurate global machine learning force fields for molecules with hundreds of atoms. Science Advances 9, eadf0873 (2023).
- Wang T, He X, Li M, Shao B, Liu TY. AIMD-Chig: Exploring the conformational space of a 166-atom protein Chignolin with ab initio molecular dynamics. Sci Data 10, 549 (2023).