Training Data, Models, and Moldens for CEONet
收藏Zenodo2025-04-29 更新2026-05-26 收录
下载链接：
https://zenodo.org/doi/10.5281/zenodo.15273310
下载链接
链接失效反馈官方服务：
资源简介：
Training data for CEONet
Three datasets were used to train the CEONet model. Each dataset contains .pt files of molecular geometries, molecular orbital coefficients, and molecular orbital energies. The .pt files are named using 'type_occupancy_size' format, where 'occ' and 'virt' indicate 'occupied' and 'virtual' orbitals, respectively.
Specifically, the three datasets are:
- sto3g: QM9 dataset with molecular orbital properties calculated at HF/STO-3G level of theory
- qh9: QM9 dataset with molecular orbital properties calculated at B3LYP/Def2-SVP level of theory
- tm: Transition metal complexes dataset with moleular orbital properties calculated at HF/STO-3G level of theory
Trained CEONet models
Models with the lowest validation mean absolute errors (MAEs) are in included in best_models.tar file
Molden Files
The hfmoldens .zip files contain the moldens of STO-3G QM9 and TMQM computed for the paper. Tmconstructed.zip contains the moldens of the constructed orbitals and tmconstructed_h5 contains the .h5 data of these orbitals that can be fed into CEONet
提供机构：
Zenodo
创建时间：
2025-04-25
