model_name: AlphaNet-v1-OAM model_key: alphanet-v1-oam model_version: v1 date_added: '2025-05-12' date_published: '2025-05-12' authors: - name: Bangchen Yin affiliation: Tsinghua University email: yinbc24@mails.tsinghua.edu.cn - name: Jiaao Wang email: wangjiaao0720@utexas.edu affiliation: University of Texas at Austin corresponding: true - name: Weitao Du affiliation: DAMO Academy, Alibaba Inc email: duweitao.dwt@alibaba-inc.com - name: Yuanqi Du affiliation: Cornell University email: yd392@cornell.edu corresponding: true - name: Chenru Duan affiliation: Deep Principle, Inc email: duanchenru@deepprinciple.com corresponding: true - name: Carla P. Gomes affiliation: Cornell University email: gomes@cs.cornell.edu - name: Graeme Henkelman corresponding: true affiliation: The University of Texas at Austin email: henkelman@utexas.edu - name: Hai Xiao affiliation: Tsinghua University email: haixiao@tsinghua.edu.cn corresponding: true trained_by: - name: Bangchen Yin affiliation: Tsinghua University email: yinbc24@mails.tsinghua.edu.cn repo: https://github.com/zmyybc/AlphaNet url: https://github.com/zmyybc/AlphaNet pr_url: https://github.com/janosh/matbench-discovery/pull/255 checkpoint_url: https://figshare.com/ndownloader/files/53851139 paper: https://arxiv.org/abs/2501.07155 doi: https://doi.org/10.48550/arXiv.2501.07155 openness: OSOD train_task: S2EFS test_task: IS2RE-SR targets: EFS_G model_type: UIP model_params: 4653577 trained_for_benchmark: false n_estimators: 1 license: code: GPL-3.0 code_url: https://github.com/zmyybc/AlphaNet/blob/main/LICENSE checkpoint: GPL-3.0 checkpoint_url: https://github.com/zmyybc/AlphaNet/blob/main/LICENSE training_cost: missing hyperparams: max_force: 0.03 max_steps: 500 ase_optimizer: FIRE cell_filter: FrechetCellFilter optimizer: Adam loss: MAE loss_weights: energy: 4.0 force: 100.0 stress: 10 batch_size: 256 #(16 per device) initial_learning_rate: 0.0002 learning_rate_schedule: StepLR(decay_steps=10000, decay_ratio = 0.93) epochs: 4 #2 epochs on OMAT24 (initial lr 0.0002), 2 epochs on sAlex and MPtrj (initial lr 0.00005) n_layers: 4 n_hidden_channels: 176 n_bessel_basis: 8 n_heads: 24 graph_construction_radius: 5.0 # Å max_neighbors: 50 requirements: torch: 2.5.1 torch-geometric: 2.6.1 torch-scatter: 2.1.2+pt25cu121 ase: 3.24.0 pymatgen: 2024.6.10 numpy: 1.26.4 training_set: ['OMat24', 'sAlex', 'MPtrj'] #train/valid split: 0.95/0.05 notes: Description: | AlphaNet is a local frame-based equivariant model designed to tackle the challenges of achieving both accurate and efficient simulations for atomistic systems. AlphaNet enhances computational efficiency and accuracy by leveraging the local geometric structures of atomic environments through the construction of equivariant local frames and learnable frame transitions. # We changed the RBF function and used a small size model in this version. metrics: phonons: kappa_103: κ_SRME: 0.6435 pred_file: models/alphanet/oam/2025-04-20-oam-kappa-103-FIRE-dist=0.03-fmax=1e-4-symprec=1e-5.json.gz pred_file_url: https://figshare.com/files/56871824 κ_SRE: 0.4914 geo_opt: pred_file: models/alphanet/oam/2025-07-24-wbm-oam-IS2RE.json.gz pred_file_url: https://figshare.com/files/56871812 struct_col: alphanet_structure symprec=1e-5: rmsd: 0.0793 # unitless n_sym_ops_mae: 10.0801 # unitless symmetry_decrease: 0.8505 # fraction symmetry_match: 0.1416 # fraction symmetry_increase: 0.0067 # fraction n_structures: 256963 # count analysis_file: models/alphanet/oam/2025-07-24-wbm-oam-IS2RE-symprec=1e-5-moyo=0.4.4.csv.gz analysis_file_url: https://figshare.com/files/56871815 symprec=1e-2: rmsd: 0.0793 # unitless n_sym_ops_mae: 2.6275 # unitless symmetry_decrease: 0.1446 # fraction symmetry_match: 0.7431 # fraction symmetry_increase: 0.1037 # fraction n_structures: 256963 # count analysis_file: models/alphanet/oam/2025-07-24-wbm-oam-IS2RE-symprec=1e-2-moyo=0.4.4.csv.gz analysis_file_url: https://figshare.com/files/56871818 discovery: pred_file: models/alphanet/oam/2025-07-24-wbm-oam-IS2RE.csv.gz pred_file_url: https://figshare.com/files/56871809 pred_col: e_form_per_atom_alphanet full_test_set: F1: 0.883 # fraction DAF: 5.0 # dimensionless Precision: 0.858 # fraction Recall: 0.91 # fraction Accuracy: 0.959 # fraction TPR: 0.91 # fraction FPR: 0.031 # fraction TNR: 0.969 # fraction FNR: 0.09 # fraction TP: 40130.0 # count FP: 6645.0 # count TN: 206226.0 # count FN: 3962.0 # count MAE: 0.023 # eV/atom RMSE: 0.075 # eV/atom R2: 0.827 # dimensionless missing_preds: 2 # count unique_prototypes: F1: 0.901 # fraction DAF: 5.747 # dimensionless Precision: 0.879 # fraction Recall: 0.924 # fraction Accuracy: 0.968 # fraction TPR: 0.924 # fraction FPR: 0.023 # fraction TNR: 0.977 # fraction FNR: 0.076 # fraction TP: 30848.0 # count FP: 4262.0 # count TN: 177852.0 # count FN: 2526.0 # count MAE: 0.024 # eV/atom RMSE: 0.076 # eV/atom R2: 0.831 # dimensionless missing_preds: 0 # count most_stable_10k: F1: 0.982 # fraction DAF: 6.312 # dimensionless Precision: 0.965 # fraction Recall: 1.0 # fraction Accuracy: 0.965 # fraction TPR: 1.0 # fraction FPR: 1.0 # fraction TNR: 0.0 # fraction FNR: 0.0 # fraction TP: 9650.0 # count FP: 350.0 # count TN: 0.0 # count FN: 0.0 # count MAE: 0.02 # eV/atom RMSE: 0.074 # eV/atom R2: 0.882 # dimensionless missing_preds: 0 # count diatomics: pred_file_url: https://figshare.com/ndownloader/files/54490370