Skip to content

Latest commit

 

History

History
59 lines (49 loc) · 3.37 KB

MODELS.md

File metadata and controls

59 lines (49 loc) · 3.37 KB

Released Models

Model Name Link MD5 Hash Matbench Discovery D3 Corrections Notes
orb-v1 link 92897eda08609425ee001955c7885139 Yes No Full dataset pretraining, MPTraj + Alexandria finetuning
orb-mptraj-only-v1 link ff42a2bc1e1f50b5f3ee2a20b83cf3a2 Yes No MPTraj pretraining and finetuning only
orb-d3-v1 link 470c7d3482ead3bc97cd4b46382d5e47 No Yes Full dataset pretraining, MPTraj + Alexandria finetuning, integrated D3 corrections
orb-d3-sm-v1 link 64fe91603e46ad5fa695525e3f1e9397 No Yes First 10 layers of a pretrained model finetuned on mptrj + alexandria with D3 corrections
orb-d3-xs-v1 link 79d042f9f16c4407795426a75498fbb7 No Yes First 5 layers of a pretrained model finetuned on mptrj + alexandria with D3 corrections

Matbench Discovery Results

orb-v1: Full dataset pretraining, MPtraj + Alexandria finetuning

                      orb          10k         unique
F1              0.846577     0.988213       0.867282
DAF             5.394101     6.389021       6.015771
Precision       0.898971     0.976700       0.919641
Recall          0.799953     1.000000       0.820563
Accuracy        0.951678     0.976700       0.961608
TPR             0.799953     1.000000       0.820563
FPR             0.017979     1.000000       0.012939
TNR             0.982021     0.000000       0.987061
FNR             0.200047     0.000000       0.179437
TP          34258.000000  9767.000000   27031.000000
FP           3850.000000   233.000000    2362.000000
TN         210288.000000     0.000000  180184.000000
FN           8567.000000     0.000000    5911.000000
MAE             0.030884     0.019012       0.030589
RMSE            0.080986     0.064470       0.079003
R2              0.798803     0.907903       0.815941

orb-mptraj-only-v1: MPTraj pretraining, MPTraj finetuning

                     orb          10k         unique
F1              0.752143     0.963193       0.761336
DAF             4.267540     6.076994       4.667345
Precision       0.711221     0.929000       0.713505
Recall          0.798062     1.000000       0.816040
Accuracy        0.912341     0.929000       0.921787
TPR             0.798062     1.000000       0.816040
FPR             0.064804     1.000000       0.059130
TNR             0.935196     0.000000       0.940870
FNR             0.201938     0.000000       0.183960
TP          34177.000000  9290.000000   26882.000000
FP          13877.000000   710.000000   10794.000000
TN         200261.000000     0.000000  171752.000000
FN           8648.000000     0.000000    6060.000000
MAE             0.044745     0.040998       0.046230
RMSE            0.093426     0.102950       0.093919
R2              0.732243     0.780546       0.739879