model_name: MACE-MP-0
model_key: mace-mp-0
model_version: v0.3.5
date_added: '2023-07-14'
date_published: '2022-05-13'
authors:
  - name: Ilyes Batatia
    affiliation: University of Cambridge
    email: ilyes.batatia@ens-paris-saclay.fr
    orcid: https://orcid.org/0000-0001-6915-9851
  - name: David P Kovacs
    affiliation: University of Cambridge
    orcid: https://orcid.org/0000-0002-0854-2635
  - name: Gregor Simm
    affiliation: University of Cambridge
    orcid: https://orcid.org/0000-0001-6815-352X
  - name: Christoph Ortner
    affiliation: University of Cambridge
    orcid: https://orcid.org/0000-0003-1498-8120
  - name: Gabor Csanyi
    affiliation: University of Cambridge
    orcid: https://orcid.org/0000-0002-8180-2034
trained_by:
  - name: Philipp Benner
    affiliation: German Federal Institute of Materials Research and Testing (BAM)
    orcid: https://orcid.org/0000-0002-0912-8137
    github: https://github.com/pbenner
  - name: Yuan Chiang
    affiliation: Lawrence Berkeley National Laboratory
    orcid: https://orcid.org/0000-0002-4017-7084
    github: https://github.com/chiang-yuan

repo: https://github.com/ACEsuit/mace
doi: https://doi.org/10.48550/arXiv.2401.00096
paper: https://arxiv.org/abs/2401.00096
pypi: https://pypi.org/project/mace-torch
pr_url: https://github.com/janosh/matbench-discovery/pull/48
# checkpoint URL copied from https://github.com/ACEsuit/mace-foundations/releases/tag/mace_mp_0
checkpoint_url: https://github.com/ACEsuit/mace-foundations/releases/download/mace_mp_0/2023-12-03-mace-128-L1_epoch-199.model

license:
  code: MIT
  code_url: https://github.com/ACEsuit/mace/blob/b0fa4ef7c/LICENSE.md
  checkpoint: MIT
  checkpoint_url: https://github.com/ACEsuit/mace-foundations/blob/1ff8786eb/LICENSE

requirements:
  mace-torch: 0.3.3
  torch: 2.0.1
  ase: 3.22.1
  pymatgen: 2023.7.14
  numpy: 1.25.0

openness: OSOD
trained_for_benchmark: true
train_task: S2EFS
test_task: IS2RE-SR
targets: EFS_G
model_type: UIP
# model_params: 2_026_624 # 2023-09-03-mace-yuan-mptrj-slower-14-lr-13_run-3
# model_params: 15_847_440 # 2023-10-29-mace-pbenner-mptrj-no-conditional-loss
model_params: 4_688_656 # 2023-12-03-mace-128-L1: https://tinyurl.com/y7uhwpje
n_estimators: 1

training_set: [MPtrj]
training_cost: missing

hyperparams:
  max_force: 0.05
  max_steps: 500
  ase_optimizer: FIRE
  cell_filter: FrechetCellFilter
  graph_construction_radius: 6.0 # Å
  max_neighbors: .inf

notes:
  Description: |
    MACE is a higher-order equivariant message-passing neural network for fast and accurate force fields.
  Training: The Matbench Discovery submission uses the same MPtrj-trained 'medium' checkpoint used for all analysis in the MACE-MP preprint.

metrics:
  phonons:
    kappa_103:
      κ_SRME: 0.6823
      pred_file: models/mace/mace-mp-0/2024-11-09-kappa-103-FIRE-dist=0.01-fmax=1e-4-symprec=1e-5.json.gz
      pred_file_url: https://figshare.com/files/52134872
      κ_SRE: 0.471
  geo_opt:
    pred_file: models/mace/mace-mp-0/2023-12-11-wbm-IS2RE-FIRE.jsonl.gz
    pred_file_url: https://figshare.com/files/57751033
    struct_col: mace_structure
    symprec=1e-5:
      rmsd: 0.0915 # unitless
      n_sym_ops_mae: 1.838 # unitless
      symmetry_decrease: 0.0335 # fraction
      symmetry_match: 0.7385 # fraction
      symmetry_increase: 0.2239 # fraction
      n_structures: 249254 # count
      analysis_file: models/mace/mace-mp-0/2023-12-11-wbm-IS2RE-FIRE-symprec=1e-5-moyo=0.4.2.csv.gz
      analysis_file_url: https://figshare.com/files/53504606
    symprec=1e-2:
      rmsd: 0.0915 # unitless
      n_sym_ops_mae: 1.8464 # unitless
      symmetry_decrease: 0.0599 # fraction
      symmetry_match: 0.8112 # fraction
      symmetry_increase: 0.122 # fraction
      n_structures: 249254 # count
      analysis_file: models/mace/mace-mp-0/2023-12-11-wbm-IS2RE-FIRE-symprec=1e-2-moyo=0.4.2.csv.gz
      analysis_file_url: https://figshare.com/files/53504657
  discovery:
    pred_file: models/mace/mace-mp-0/2023-12-11-wbm-IS2RE-FIRE.csv.gz
    pred_file_url: https://figshare.com/files/52057538
    pred_col: e_form_per_atom_mace
    full_test_set:
      F1: 0.668 # fraction
      DAF: 3.4 # dimensionless
      Precision: 0.583 # fraction
      Recall: 0.781 # fraction
      Accuracy: 0.867 # fraction
      TPR: 0.781 # fraction
      FPR: 0.115 # fraction
      TNR: 0.885 # fraction
      FNR: 0.219 # fraction
      TP: 34420.0 # count
      FP: 24576.0 # count
      TN: 188295.0 # count
      FN: 9672.0 # count
      MAE: 0.055 # eV/atom
      RMSE: 0.099 # eV/atom
      R2: 0.698 # dimensionless
      missing_preds: 38 # count
    most_stable_10k:
      F1: 0.888 # fraction
      DAF: 5.221 # dimensionless
      Precision: 0.798 # fraction
      Recall: 1.0 # fraction
      Accuracy: 0.798 # fraction
      TPR: 1.0 # fraction
      FPR: 1.0 # fraction
      TNR: 0.0 # fraction
      FNR: 0.0 # fraction
      TP: 7982.0 # count
      FP: 2018.0 # count
      TN: 0.0 # count
      FN: 0.0 # count
      MAE: 0.087 # eV/atom
      RMSE: 0.165 # eV/atom
      R2: 0.508 # dimensionless
      missing_preds: 0 # count
    unique_prototypes:
      F1: 0.669 # fraction
      DAF: 3.777 # dimensionless
      Precision: 0.577 # fraction
      Recall: 0.796 # fraction
      Accuracy: 0.878 # fraction
      TPR: 0.796 # fraction
      FPR: 0.107 # fraction
      TNR: 0.893 # fraction
      FNR: 0.204 # fraction
      TP: 26582.0 # count
      FP: 19457.0 # count
      TN: 162657.0 # count
      FN: 6792.0 # count
      MAE: 0.057 # eV/atom
      RMSE: 0.101 # eV/atom
      R2: 0.697 # dimensionless
      missing_preds: 34 # count
  diatomics:
    pred_file: models/mace/mace-mp-0/2025-02-13-diatomics.json.gz
    pred_file_url: https://figshare.com/files/52449434