model_name: eSEN-30M-MP
model_key: eSEN-30m-mp
model_version: v2025.03.17
date_added: '2025-03-17'
date_published: '2025-03-17'
authors:
  - name: Xiang Fu
    affiliation: FAIR at Meta
    email: xiangfu@meta.com
    orcid: https://orcid.org/0000-0001-7480-6312
    github: https://github.com/kyonofx
    corresponding: true
  - name: Brandon M. Wood
    affiliation: FAIR at Meta
  - name: Luis Barroso-Luque
    affiliation: FAIR at Meta
  - name: Daniel S. Levine
    affiliation: FAIR at Meta
  - name: Meng Gao
    affiliation: FAIR at Meta
  - name: Misko Dzamba
    affiliation: FAIR at Meta
  - name: C. Lawrence Zitnick
    affiliation: FAIR at Meta
    email: zitnick@meta.com
    corresponding: true

repo: https://github.com/FAIR-Chem/fairchem
doi: https://doi.org/10.48550/arXiv.2502.12147
paper: https://arxiv.org/abs/2502.12147
url: https://huggingface.co/facebook/OMAT24#model-checkpoints
pypi: https://pypi.org/project/fairchem-core
pr_url: https://github.com/janosh/matbench-discovery/pull/226
checkpoint_url: https://huggingface.co/fairchem/OMAT24/blob/main/esen_30m_oam.pt # taken from https://huggingface.co/facebook/OMAT24#models-trained-on-omat-mptrj-and-salexandria

license:
  code: MIT
  code_url: https://github.com/FAIR-Chem/fairchem/blob/aa160789e1/LICENSE.md
  checkpoint: Meta Research
  checkpoint_url: https://huggingface.co/facebook/OMAT24/blob/main/LICENSE

requirements:
  fairchem-core: 1.7.0

openness: OSOD
trained_for_benchmark: false
train_task: S2EFS
test_task: IS2RE-SR
targets: EFS_G
model_type: UIP
model_params: 30_085_121
n_estimators: 1

training_set: [MPtrj]
training_cost: missing

hyperparams:
  max_force: 0.02
  max_steps: 500
  ase_optimizer: FIRE
  cell_filter: FrechetCellFilter
  graph_construction_radius: 6.0 # Å, from table 5 in arXiv:2502.12147
  max_neighbors: .inf # see sec. 5.3 in arXiv:2502.12147

notes:
  Description: |
    equivariant Smooth Energy Network (eSEN), described in https://arxiv.org/abs/2502.12147.
  Training: |
    Training was done by: (1) 70-epoch direct pre-training on MPtrj with DeNS; (2) 30-epoch conservative fine-tuning on MPtrj without DeNS.

metrics:
  phonons:
    kappa_103:
      κ_SRME: 0.3398
      pred_file: models/eSEN/eSEN-30m-mp/2025-03-18-kappa-103-FIRE-dist=0.03-fmax=1e-4-symprec=1e-5.json.gz
      pred_file_url: https://figshare.com/files/53090588
      κ_SRE: 0.1854
  geo_opt:
    pred_file: models/eSEN/eSEN-30m-mp/2025-03-17-wbm-geo-opt.jsonl.gz
    pred_file_url: https://figshare.com/files/53153477
    struct_col: pred_structure_esen
    symprec=1e-5:
      rmsd: 0.0752 # unitless
      n_sym_ops_mae: 3.3144 # unitless
      symmetry_decrease: 0.4281 # fraction
      symmetry_match: 0.3859 # fraction
      symmetry_increase: 0.1282 # fraction
      n_structures: 256963 # count
      analysis_file: models/eSEN/eSEN-30m-mp/2025-03-17-wbm-geo-opt-symprec=1e-5-moyo=0.4.2.csv.gz
      analysis_file_url: https://figshare.com/files/53504555
    symprec=1e-2:
      rmsd: 0.0752 # unitless
      n_sym_ops_mae: 2.521 # unitless
      symmetry_decrease: 0.2604 # fraction
      symmetry_match: 0.6106 # fraction
      symmetry_increase: 0.098 # fraction
      n_structures: 256963 # count
      analysis_file: models/eSEN/eSEN-30m-mp/2025-03-17-wbm-geo-opt-symprec=1e-2-moyo=0.4.2.csv.gz
      analysis_file_url: https://figshare.com/files/53504558
  discovery:
    pred_file: models/eSEN/eSEN-30m-mp/2025-03-17-wbm-IS2RE.csv.gz
    pred_file_url: https://figshare.com/files/53054366
    pred_col: pred_e_form_per_atom_esen
    full_test_set:
      F1: 0.819 # fraction
      DAF: 4.616 # dimensionless
      Precision: 0.792 # fraction
      Recall: 0.849 # fraction
      Accuracy: 0.936 # fraction
      TPR: 0.849 # fraction
      FPR: 0.046 # fraction
      TNR: 0.954 # fraction
      FNR: 0.151 # fraction
      TP: 37419.0 # count
      FP: 9825.0 # count
      TN: 203046.0 # count
      FN: 6673.0 # count
      MAE: 0.032 # eV/atom
      RMSE: 0.077 # eV/atom
      R2: 0.818 # dimensionless
      missing_preds: 2 # count
    unique_prototypes:
      F1: 0.831 # fraction
      DAF: 5.26 # dimensionless
      Precision: 0.804 # fraction
      Recall: 0.861 # fraction
      Accuracy: 0.946 # fraction
      TPR: 0.861 # fraction
      FPR: 0.038 # fraction
      TNR: 0.962 # fraction
      FNR: 0.139 # fraction
      TP: 28722.0 # count
      FP: 7000.0 # count
      TN: 175114.0 # count
      FN: 4652.0 # count
      MAE: 0.033 # eV/atom
      RMSE: 0.078 # eV/atom
      R2: 0.822 # dimensionless
      missing_preds: 0 # count
    most_stable_10k:
      F1: 0.978 # fraction
      DAF: 6.261 # dimensionless
      Precision: 0.957 # fraction
      Recall: 1.0 # fraction
      Accuracy: 0.957 # fraction
      TPR: 1.0 # fraction
      FPR: 1.0 # fraction
      TNR: 0.0 # fraction
      FNR: 0.0 # fraction
      TP: 9572.0 # count
      FP: 428.0 # count
      TN: 0.0 # count
      FN: 0.0 # count
      MAE: 0.035 # eV/atom
      RMSE: 0.111 # eV/atom
      R2: 0.755 # dimensionless
      missing_preds: 0 # count