model_name: SevenNet-Omni-i12 model_key: sevennet-omni-i12 model_version: v0.12.0 date_added: '2026-01-12' date_published: '2026-01-12' authors: - name: Jaesun Kim affiliation: Seoul National University email: sunny990912@snu.ac.kr orcid: https://orcid.org/0009-0000-6646-1318 - name: Jinmu You affiliation: Seoul National University email: alphalm4@snu.ac.kr orcid: https://orcid.org/0009-0003-7152-7774 - name: Yutack Park affiliation: Seoul National University email: parkyutack@snu.ac.kr orcid: https://orcid.org/0009-0008-8690-935X - name: Suyeon Ju affiliation: Seoul National University email: cindy2069@snu.ac.kr orcid: https://orcid.org/0000-0001-7033-8769 - name: Haekwan Jeon affiliation: Seoul National University email: haekwan98@snu.ac.kr orcid: https://orcid.org/0009-0000-3088-7499 - name: Seungwu Han affiliation: Seoul National University, Korea Institute for Advanced Study email: hansw@snu.ac.kr orcid: https://orcid.org/0000-0003-3958-0922 corresponding: true trained_by: - name: Jinmu You affiliation: Seoul National University email: alphalm4@snu.ac.kr orcid: https://orcid.org/0009-0003-7152-7774 repo: https://github.com/MDIL-SNU/SevenNet url: https://figshare.com/account/articles/31054558 doi: https://doi.org/10.48550/arXiv.2510.11241 paper: https://arxiv.org/abs/2510.11241 pypi: https://pypi.org/project/sevenn pr_url: https://github.com/janosh/matbench-discovery/pull/305 checkpoint_url: https://figshare.com/files/60977863 license: code: MIT code_url: https://github.com/MDIL-SNU/SevenNet/blob/main/LICENSE checkpoint: MIT checkpoint_url: https://figshare.com/files/60977863 openness: OSOD trained_for_benchmark: false train_task: S2EFS test_task: IS2RE-SR targets: EFS_G model_type: UIP model_params: 54_921_363 n_estimators: 1 hyperparams: max_force: 0.02 max_steps: 800 ase_optimizer: FIRE cell_filter: FrechetCellFilter optimizer: Adam loss: MAE/L2MAE/L2MAE loss_weights: energy: 1.0 force: 1.0 stress: 0.0005 batch_size: 256 initial_learning_rate: 0.0001 learning_rate_schedule: onecyclelr - max_lr=0.0001, pct_start=0.05, anneal_strategy=cos, div_factor=25, final_div_factor=1e4 epochs: 2 n_layers: 12 n_features: - 128x0e - 128x0e+64x1o+32x2e+32x3o - 128x0e+64x1o+32x2e+32x3o - 128x0e+64x1o+32x2e+32x3o - 128x0e+64x1o+32x2e+32x3o - 128x0e+64x1o+32x2e+32x3o - 128x0e+64x1o+32x2e+32x3o - 128x0e+64x1o+32x2e+32x3o - 128x0e+64x1o+32x2e+32x3o - 128x0e+64x1o+32x2e+32x3o - 128x0e+64x1o+32x2e+32x3o - 128x0e+64x1o+32x2e+32x3o - 128x0e n_radial_bessel_basis: 8 graph_construction_radius: 6.0 # Å, from SevenNet-mf-ompa/hyperparams.yaml max_neighbors: .inf sph_harmonics_l_max: 3 requirements: torch: 2.7.0 torch-geometric: 2.6.1 ase: 3.23.0 pymatgen: 2025.10.7 numpy: 1.26.4 training_set: [COSMOSDataset] # # curriculum learning: ompa (260) -> intermediate (130) -> omni (410) training_cost: Nvidia H200 GPUs: { amount: 8, hours: 800, cost: 6400 } notes: Description: | SevenNet is a graph neural network interatomic potential package that supports parallel molecular dynamics simulations. The SevenNet-Omni model employs a multi-task training strategy that jointly optimizes universal and task-specific parameters via selective regularization and domain-bridging strategies, enabling robust transferability across molecules, bulk crystals, and surfaces. Trained on 15 open datasets spanning molecular, inorganic, and interfacial chemistries, SevenNet-Omni achieves state-of-the-art cross-domain accuracy while maintaining high in-domain fidelity. metrics: phonons: kappa_103: pred_file: models/sevennet/sevennet-omni-i12/2026-01-12-kappa-103-FIRE-dist=0.03-fmax=1e-4-symprec=1e-05.json.gz pred_file_url: https://figshare.com/files/60977131 κ_SRME: 0.1917 κ_SRE: 0.0963 geo_opt: pred_file: models/sevennet/sevennet-omni-i12/2026-01-12-wbm-IS2RE-FIRE.jsonl.gz pred_file_url: https://figshare.com/files/60979039 struct_col: sevennet_structure symprec=1e-2: rmsd: 0.0618 # unitless n_sym_ops_mae: 1.7097 # unitless symmetry_decrease: 0.0469 # fraction symmetry_match: 0.8187 # fraction symmetry_increase: 0.1272 # fraction n_structures: 256963 # count analysis_file: models/sevennet/sevennet-omni-i12/2026-01-12-wbm-IS2RE-FIRE-symprec=1e-2-moyo=0.7.3.csv.gz analysis_file_url: https://figshare.com/files/60979033 symprec=1e-5: rmsd: 0.0618 # unitless n_sym_ops_mae: 2.0291 # unitless symmetry_decrease: 0.0439 # fraction symmetry_match: 0.7056 # fraction symmetry_increase: 0.2454 # fraction n_structures: 256963 # count analysis_file: models/sevennet/sevennet-omni-i12/2026-01-12-wbm-IS2RE-FIRE-symprec=1e-5-moyo=0.7.3.csv.gz analysis_file_url: https://figshare.com/files/60979036 discovery: pred_file: models/sevennet/sevennet-omni-i12/2026-01-12-wbm-IS2RE.csv.gz pred_file_url: https://figshare.com/files/60977137 pred_col: e_form_per_atom_sevennet full_test_set: F1: 0.887 # fraction DAF: 5.203 # dimensionless Precision: 0.893 # fraction Recall: 0.882 # fraction Accuracy: 0.962 # fraction TPR: 0.882 # fraction FPR: 0.022 # fraction TNR: 0.978 # fraction FNR: 0.118 # fraction TP: 38875.0 # count FP: 4668.0 # count TN: 208203.0 # count FN: 5217.0 # count MAE: 0.02 # eV/atom RMSE: 0.067 # eV/atom R2: 0.862 # dimensionless missing_preds: 2 # count unique_prototypes: F1: 0.906 # fraction DAF: 5.954 # dimensionless Precision: 0.91 # fraction Recall: 0.901 # fraction Accuracy: 0.971 # fraction TPR: 0.901 # fraction FPR: 0.016 # fraction TNR: 0.984 # fraction FNR: 0.099 # fraction TP: 30084.0 # count FP: 2969.0 # count TN: 179145.0 # count FN: 3290.0 # count MAE: 0.021 # eV/atom RMSE: 0.067 # eV/atom R2: 0.868 # dimensionless missing_preds: 0 # count most_stable_10k: F1: 0.985 # fraction DAF: 6.352 # dimensionless Precision: 0.971 # fraction Recall: 1.0 # fraction Accuracy: 0.971 # fraction TPR: 1.0 # fraction FPR: 1.0 # fraction TNR: 0.0 # fraction FNR: 0.0 # fraction TP: 9711.0 # count FP: 289.0 # count TN: 0.0 # count FN: 0.0 # count MAE: 0.019 # eV/atom RMSE: 0.078 # eV/atom R2: 0.868 # dimensionless missing_preds: 0 # count