model_name: CHGNet model_key: chgnet-0.3.0 model_version: v0.3.0 date_added: '2023-03-03' date_published: '2023-03-01' authors: - name: Bowen Deng affiliation: UC Berkeley orcid: https://orcid.org/0000-0002-5720-5299 - name: Peichen Zhong affiliation: UC Berkeley orcid: https://orcid.org/0000-0003-1921-1628 email: zhongpc@berkeley.edu - name: KyuJung Jun affiliation: UC Berkeley orcid: https://orcid.org/0000-0003-1974-028X - name: Janosh Riebesell affiliation: University of Cambridge, Lawrence Berkeley National Laboratory email: janosh.riebesell@gmail.com orcid: https://orcid.org/0000-0001-5233-3462 - name: Kevin Han affiliation: UC Berkeley orcid: https://orcid.org/0000-0002-4028-2108 - name: Christopher J. Bartel affiliation: University of Minnesota orcid: https://orcid.org/0000-0002-5198-5036 - name: Gerbrand Ceder affiliation: UC Berkeley orcid: https://orcid.org/0000-0001-9275-3605 email: gceder@berkeley.edu repo: https://github.com/CederGroupHub/chgnet doi: https://doi.org/10.48550/arXiv.2302.14231 paper: https://arxiv.org/abs/2302.14231 pypi: https://pypi.org/project/chgnet pr_url: https://github.com/janosh/matbench-discovery/pull/85 # checkpoint URL copied from https://github.com/CederGroupHub/chgnet/blob/d55f185199fddc9/chgnet/pretrained/0.3.0/chgnet_0.3.0_e29f68s314m37.pth.tar checkpoint_url: https://github.com/CederGroupHub/chgnet/raw/refs/heads/main/chgnet/pretrained/0.3.0/chgnet_0.3.0_e29f68s314m37.pth.tar license: code: BSD-3-Clause code_url: https://github.com/CederGroupHub/chgnet/blob/d55f185199f/LICENSE checkpoint: BSD-3-Clause checkpoint_url: https://github.com/CederGroupHub/chgnet/blob/d55f185199f/LICENSE requirements: torch: 1.11.0 ase: 3.22.0 pymatgen: 2022.10.22 numpy: 1.24.0 openness: OSOD trained_for_benchmark: false train_task: S2EFSM test_task: IS2RE-SR targets: EFS_GM model_type: UIP model_params: 412_525 n_estimators: 1 training_set: [MPtrj] training_cost: missing hyperparams: ase_optimizer: FIRE cell_filter: FrechetCellFilter max_steps: 500 max_force: 0.05 # eV/Å graph_construction_radius: 5.0 # Å, from sec. B. Model design in arXiv:2302.14231 three_body_cutoff: 3.0 # Å max_neighbors: .inf notes: Description: | The Crystal Hamiltonian Graph Neural Network (CHGNet) is a universal GNN-based interatomic potential trained on energies, forces, stresses and magnetic moments from the MP trajectory dataset containing ∼1.5 million inorganic structures. ![CHGNet Pipeline](https://user-images.githubusercontent.com/30958850/222924937-1d09bbce-ee18-4b19-8061-ec689cd15887.svg) Training: Using pre-trained model with 400,438 params released with paper. The MPtrj data set used to train CHGNet was pulled in September 2022 from the at-the-time latest [v2021.11.10 MP release](https://docs.materialsproject.org/changes/database-versions#v2021.11.10) (see [description for construction of MPtrj](https://github.com/CederGroupHub/chgnet/blob/7c21a9488/examples/QueryMPtrj.md)). The CHGNet authors' MPtrj data set has since been used to train other models like [MACE-MP](https://arxiv.org/abs/2401.00096) and [Equiformer v1/2](https://github.com/pbenner/equitrain). Corrections: Unlike e.g. [M3GNet](/models/m3gnet) which predicts raw DFT energies, CHGNet targets include MP2020 corrections. Hence no need to correct again. metrics: phonons: kappa_103: κ_SRME: 2.0 pred_file: models/chgnet/chgnet-0.3.0/2024-11-09-kappa-103-FIRE-fmax=1e-4-symprec=1e-5.json.gz pred_file_url: https://figshare.com/files/52134857 κ_SRE: 2.0 geo_opt: pred_file: models/chgnet/chgnet-0.3.0/2023-12-21-wbm-geo-opt.jsonl.gz pred_file_url: https://figshare.com/files/53153459 struct_col: chgnet_structure symprec=1e-5: rmsd: 0.0949 # unitless n_sym_ops_mae: 3.0863 # unitless symmetry_decrease: 0.2241 # fraction symmetry_match: 0.6062 # fraction symmetry_increase: 0.1601 # fraction n_structures: 256963 # count analysis_file: models/chgnet/chgnet-0.3.0/2023-12-21-wbm-geo-opt-symprec=1e-5-moyo=0.4.2.csv.gz analysis_file_url: https://figshare.com/files/53504513 symprec=1e-2: rmsd: 0.0949 # unitless n_sym_ops_mae: 2.0181 # unitless symmetry_decrease: 0.0922 # fraction symmetry_match: 0.7921 # fraction symmetry_increase: 0.1076 # fraction n_structures: 256963 # count analysis_file: models/chgnet/chgnet-0.3.0/2023-12-21-wbm-geo-opt-symprec=1e-2-moyo=0.4.2.csv.gz analysis_file_url: https://figshare.com/files/53504516 discovery: pred_file: models/chgnet/chgnet-0.3.0/2023-12-21-wbm-IS2RE.csv.gz pred_file_url: https://figshare.com/files/52057526 pred_col: e_form_per_atom_chgnet full_test_set: F1: 0.612 # fraction DAF: 3.038 # dimensionless Precision: 0.521 # fraction Recall: 0.74 # fraction Accuracy: 0.839 # fraction TPR: 0.74 # fraction FPR: 0.141 # fraction TNR: 0.859 # fraction FNR: 0.26 # fraction TP: 32642.0 # count FP: 29979.0 # count TN: 182892.0 # count FN: 11450.0 # count MAE: 0.061 # eV/atom RMSE: 0.1 # eV/atom R2: 0.69 # dimensionless missing_preds: 2 # count most_stable_10k: F1: 0.92 # fraction DAF: 5.567 # dimensionless Precision: 0.851 # fraction Recall: 1.0 # fraction Accuracy: 0.851 # fraction TPR: 1.0 # fraction FPR: 1.0 # fraction TNR: 0.0 # fraction FNR: 0.0 # fraction TP: 8511.0 # count FP: 1489.0 # count TN: 0.0 # count FN: 0.0 # count MAE: 0.063 # eV/atom RMSE: 0.095 # eV/atom R2: 0.816 # dimensionless missing_preds: 0 # count unique_prototypes: F1: 0.613 # fraction DAF: 3.361 # dimensionless Precision: 0.514 # fraction Recall: 0.758 # fraction Accuracy: 0.851 # fraction TPR: 0.758 # fraction FPR: 0.132 # fraction TNR: 0.868 # fraction FNR: 0.242 # fraction TP: 25313.0 # count FP: 23955.0 # count TN: 158159.0 # count FN: 8061.0 # count MAE: 0.063 # eV/atom RMSE: 0.103 # eV/atom R2: 0.689 # dimensionless missing_preds: 0 # count