model_name: Eqnorm MPtrj model_key: eqnorm-mptrj # this should match the name of the YAML file and determines the URL /models/ on which details of the model are displayed on the website model_version: '0.1.0' date_added: '2025-05-26' date_published: '2025-05-26' authors: - name: Yuzhuo Chen affiliation: Zhejiang Lab email: chenyuzhuo@zhejianglab.org orcid: https://orcid.org/0009-0003-4044-2372 - name: Lyuwen Fu affiliation: Zhejiang Lab email: lyuwenfu@zhejianglab.org - name: Shuxiang Yang affiliation: Zhejiang Lab email: yang_shuxiang@zhejianglab.org - name: Lipeng Chen affiliation: Zhejiang Lab email: chenlp@zhejianglab.org corresponding: true trained_by: - name: Yuzhuo Chen affiliation: Zhejiang Lab email: chenyuzhuo@zhejianglab.org orcid: https://orcid.org/0009-0003-4044-2372 repo: https://github.com/yzchen08/eqnorm url: https://github.com/yzchen08/eqnorm/tree/main/eqnorm pr_url: https://github.com/janosh/matbench-discovery/pull/258 doi: https://github.com/yzchen08/eqnorm # to be released soon paper: https://github.com/yzchen08/eqnorm # to be released soon checkpoint_url: https://figshare.com/files/55429685 license: code: MIT code_url: https://github.com/yzchen08/eqnorm/blob/main/LICENSE checkpoint: MIT checkpoint_url: https://github.com/yzchen08/eqnorm/blob/main/LICENSE openness: OSOD train_task: S2EFS test_task: IS2RE-SR targets: EFS_G model_type: UIP model_params: 1310024 trained_for_benchmark: true n_estimators: 1 hyperparams: max_force: 0.02 max_steps: 500 ase_optimizer: FIRE cell_filter: FrechetCellFilter loss: Huber loss_weights: energy: 20.0 force: 20.0 stress: 320.0 optimizer: AdamW weight_decay: 0.001 clip_grad_norm: 100 ema_decay: 0.999 max_learning_rate: 0.01 min_learning_rate: 0.000001 learning_rate_schedule: warmcosine warmup_factor: 0.2 epochs: 100 batch_train: 128 n_layers: 4 num_embedding_features: 128 num_bessel_basis: 8 invariant_layers: 2 invariant_neurons: 64 poly_p: 6 graph_construction_radius: 6.0 max_neighbors: .inf irreps_hidden: 128x0e+64x1o+32x2e+32x3o irreps_sh: 1x0e+1x1o+1x2e+1x3o energy_shift: per_species energy_scale: force_rms shift_trainable: false scale_trainable: false requirements: torch: '2.2.2' torch-geometric: '2.6.1' ase: '3.24.0' vesin: '0.3.2' e3nn: '0.5.6' pymatgen: '2025.3.10' numpy: '2.2.4' wget: '3.2' training_set: [MPtrj] training_cost: Nvidia A100 GPUs: { amount: 4, hours: 500 } notes: Description: | eqnorm is a graph neural network model designed for predicting the energy, forces, and stresses of materials. The model utilizes a combination of invariant and equivariant layers to effectively capture the symmetries present in material structures. metrics: phonons: kappa_103: κ_SRME: 0.4079 pred_file: models/eqnorm/eqnorm-mptrj/2025-06-16-kappa-103-FIRE-dist=0.03-fmax=1e-4-symprec=1e-5.json.gz pred_file_url: https://figshare.com/files/55429706 κ_SRE: 0.2722 geo_opt: pred_file: models/eqnorm/eqnorm-mptrj/2025-06-16-wbm-IS2RE-FIRE.jsonl.gz pred_file_url: https://figshare.com/files/55550972 struct_col: eqnorm_structure symprec=1e-2: rmsd: 0.0837 # unitless n_sym_ops_mae: 1.7966 # unitless symmetry_decrease: 0.0589 # fraction symmetry_match: 0.8113 # fraction symmetry_increase: 0.1224 # fraction n_structures: 256963 # count analysis_file: models/eqnorm/eqnorm-mptrj/2025-06-16-wbm-IS2RE-FIRE-symprec=1e-2-moyo=0.4.2.csv.gz analysis_file_url: https://figshare.com/files/55550975 symprec=1e-5: rmsd: 0.0837 # unitless n_sym_ops_mae: 2.0441 # unitless symmetry_decrease: 0.0639 # fraction symmetry_match: 0.6983 # fraction symmetry_increase: 0.2316 # fraction n_structures: 256963 # count analysis_file: models/eqnorm/eqnorm-mptrj/2025-06-16-wbm-IS2RE-FIRE-symprec=1e-5-moyo=0.4.2.csv.gz analysis_file_url: https://figshare.com/files/55550978 discovery: pred_file: models/eqnorm/eqnorm-mptrj/2025-06-16-wbm-IS2RE.csv.gz pred_file_url: https://figshare.com/files/55546649 pred_col: e_form_per_atom_eqnorm full_test_set: F1: 0.779 # fraction DAF: 4.296 # dimensionless Precision: 0.737 # fraction Recall: 0.825 # fraction Accuracy: 0.919 # fraction TPR: 0.825 # fraction FPR: 0.061 # fraction TNR: 0.939 # fraction FNR: 0.175 # fraction TP: 36369.0 # count FP: 12963.0 # count TN: 199908.0 # count FN: 7723.0 # count MAE: 0.038 # eV/atom RMSE: 0.081 # eV/atom R2: 0.796 # dimensionless missing_preds: 2 # count unique_prototypes: F1: 0.786 # fraction DAF: 4.844 # dimensionless Precision: 0.741 # fraction Recall: 0.838 # fraction Accuracy: 0.929 # fraction TPR: 0.838 # fraction FPR: 0.054 # fraction TNR: 0.946 # fraction FNR: 0.162 # fraction TP: 27984.0 # count FP: 9804.0 # count TN: 172310.0 # count FN: 5390.0 # count MAE: 0.04 # eV/atom RMSE: 0.083 # eV/atom R2: 0.799 # dimensionless missing_preds: 0 # count most_stable_10k: F1: 0.97 # fraction DAF: 6.156 # dimensionless Precision: 0.941 # fraction Recall: 1.0 # fraction Accuracy: 0.941 # fraction TPR: 1.0 # fraction FPR: 1.0 # fraction TNR: 0.0 # fraction FNR: 0.0 # fraction TP: 9411.0 # count FP: 589.0 # count TN: 0.0 # count FN: 0.0 # count MAE: 0.039 # eV/atom RMSE: 0.105 # eV/atom R2: 0.775 # dimensionless missing_preds: 0 # count