model_name: AlphaNet-v1-MPtrj model_key: alphanet-v1-mptrj model_version: v1 date_added: '2025-03-05' date_published: '2025-03-05' 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 email: duanchenru@deepprinciple.com corresponding: true - name: Carla P. Gomes affiliation: Cornell University email: gomes@cs.cornell.edu - name: Graeme Henkelman 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/216 paper: https://arxiv.org/abs/2501.07155 doi: https://doi.org/10.48550/arXiv.2501.07155 # checkpoint page: https://github.com/zmyybc/AlphaNet/blob/243fe71cb96/README.md#pretrained-models checkpoint_url: https://figshare.com/files/52870784 license: code: GPL-3.0 code_url: https://github.com/zmyybc/AlphaNet/blob/243fe71cb/LICENSE checkpoint: CC-BY-4.0 checkpoint_url: https://figshare.com/articles/dataset/mp-0225-2_ckpt/28560176?file=52870784 openness: OSOD train_task: S2EFS test_task: IS2RE-SR targets: EFS_G model_type: UIP model_params: 16234332 trained_for_benchmark: true n_estimators: 1 status: superseded superseded_by: alphanet-v1-oam hyperparams: max_force: 0.05 max_steps: 500 ase_optimizer: FIRE cell_filter: FrechetCellFilter optimizer: Adam loss: MAE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: 128 #(8 per device) initial_learning_rate: 0.0002 learning_rate_schedule: CosineAnnealingLR(Tmax=200, lr_min = 1e-6) epochs: 200 n_layers: 6 n_hidden_channels: 256 n_radial_gaussian_basis: 256 n_heads: 24 graph_construction_radius: 6.0 # Å, from table 6 in arXiv:2501.07155 max_neighbors: .inf # max_num_neighbors_threshold seems to be unused in https://github.com/zmyybc/AlphaNet/blob/243fe71cb9/alphanet/models/graph.py#L28 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: [MPtrj] #train/valid split: 0.95/0.05 training_cost: missing 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. metrics: phonons: kappa_103: κ_SRME: 1.310 pred_file: models/alphanet/mptrj/2025-03-04-kappa-103-fire-dist=0.01-fmax=1e-4-symprec=1e-5.json.gz #find a lots of imag freqs, will look into this in the future pred_file_url: https://figshare.com/files/52869011 geo_opt: pred_file: models/alphanet/mptrj/2025-03-04-wbm-geo-opt-fire.jsonl.gz pred_file_url: https://figshare.com/files/53153453 struct_col: alphanet_structure symprec=1e-5: rmsd: 0.1067 # unitless n_sym_ops_mae: 10.1098 # unitless symmetry_decrease: 0.8597 # fraction symmetry_match: 0.1376 # fraction symmetry_increase: 0.0024 # fraction n_structures: 256963 # count analysis_file: models/alphanet/mptrj/2025-03-04-wbm-geo-opt-fire-symprec=1e-5-moyo=0.4.2.csv.gz analysis_file_url: https://figshare.com/files/53504501 symprec=1e-2: rmsd: 0.1067 # unitless n_sym_ops_mae: 8.1732 # unitless symmetry_decrease: 0.5768 # fraction symmetry_match: 0.3712 # fraction symmetry_increase: 0.0419 # fraction n_structures: 256963 # count analysis_file: models/alphanet/mptrj/2025-03-04-wbm-geo-opt-fire-symprec=1e-2-moyo=0.4.2.csv.gz analysis_file_url: https://figshare.com/files/53504504 discovery: pred_file: models/alphanet/mptrj/2025-03-04-wbm-IS2RE.csv.gz pred_file_url: https://figshare.com/files/52869026 pred_col: e_form_per_atom_alphanet full_test_set: F1: 0.789 # fraction DAF: 4.312 # dimensionless Precision: 0.74 # fraction Recall: 0.846 # fraction Accuracy: 0.923 # fraction TPR: 0.846 # fraction FPR: 0.062 # fraction TNR: 0.938 # fraction FNR: 0.154 # fraction TP: 37311.0 # count FP: 13119.0 # count TN: 199752.0 # count FN: 6781.0 # count MAE: 0.04 # eV/atom RMSE: 0.091 # eV/atom R2: 0.747 # dimensionless missing_preds: 2 # count unique_prototypes: F1: 0.799 # fraction DAF: 4.863 # dimensionless Precision: 0.743 # fraction Recall: 0.864 # fraction Accuracy: 0.933 # fraction TPR: 0.864 # fraction FPR: 0.055 # fraction TNR: 0.945 # fraction FNR: 0.136 # fraction TP: 28831.0 # count FP: 9949.0 # count TN: 172165.0 # count FN: 4543.0 # count MAE: 0.041 # eV/atom RMSE: 0.093 # eV/atom R2: 0.745 # dimensionless missing_preds: 0 # count most_stable_10k: F1: 0.96 # fraction DAF: 6.042 # dimensionless Precision: 0.924 # fraction Recall: 1.0 # fraction Accuracy: 0.924 # fraction TPR: 1.0 # fraction FPR: 1.0 # fraction TNR: 0.0 # fraction FNR: 0.0 # fraction TP: 9237.0 # count FP: 763.0 # count TN: 0.0 # count FN: 0.0 # count MAE: 0.045 # eV/atom RMSE: 0.115 # eV/atom R2: 0.746 # dimensionless missing_preds: 0 # count diatomics: pred_file: models/alphanet/2025-03-08-diatomics.json.gz pred_file_url: https://figshare.com/files/52868972