Pretrained FAIRChem models

Pretrained FAIRChem models#

This page summarizes all the pretrained models released as part of the Open Catalyst Project. All models were trained using this codebase.

Open Catalyst 2020 (OC20)#

  • All configurations for these models are available in the configs/ directory.

  • All of these models are trained on various splits of the OC20 S2EF / IS2RE datasets. For details, see https://arxiv.org/abs/2010.09990 and https://github.com/FAIR-Chem/fairchem/blob/main/DATASET.md.

  • All OC20 models are trained on adsorption energies, i.e. the DFT total energies minus the clean surface and gas phase adsorbate energies. For details on how to train models on OC20 total energies, please read the referencing section here.

S2EF models: optimized for EFwT#

Model Name

Model

Split

Download

val ID force MAE (eV / Å)

val ID EFwT

CGCNN-S2EF-OC20-200k

CGCNN

200k

checkpoint | config

0.08

0%

CGCNN-S2EF-OC20-2M

CGCNN

2M

checkpoint | config

0.0673

0.01%

CGCNN-S2EF-OC20-20M

CGCNN

20M

checkpoint | config

0.065

0%

CGCNN-S2EF-OC20-All

CGCNN

All

checkpoint | config

0.0684

0.01%

DimeNet-S2EF-OC20-200k

DimeNet

200k

checkpoint

0.0693

0.01%

DimeNet-S2EF-OC20-2M

DimeNet

2M

checkpoint

0.0576

0.02%

SchNet-S2EF-OC20-200k

SchNet

200k

checkpoint | config

0.0743

0%

SchNet-S2EF-OC20-2M

SchNet

2M

checkpoint | config

0.0737

0%

SchNet-S2EF-OC20-20M

SchNet

20M

checkpoint | config

0.0568

0.03%

SchNet-S2EF-OC20-All

SchNet

All

checkpoint | config

0.0494

0.12%

DimeNet++-S2EF-OC20-200k

DimeNet++

200k

checkpoint | config

0.0741

0%

DimeNet++-S2EF-OC20-2M

DimeNet++

2M

checkpoint | config

0.0595

0.01%

DimeNet++-S2EF-OC20-20M

DimeNet++

20M

checkpoint | config

0.0511

0.06%

DimeNet++-S2EF-OC20-All

DimeNet++

All

checkpoint | config

0.0444

0.12%

SpinConv-S2EF-OC20-2M

SpinConv

2M

checkpoint | config

0.0329

0.18%

SpinConv-S2EF-OC20-All

SpinConv

All

checkpoint | config

0.0267

1.02%

GemNet-dT-S2EF-OC20-2M

GemNet-dT

2M

checkpoint | config

0.0257

1.10%

GemNet-dT-S2EF-OC20-All

GemNet-dT

All

checkpoint | config

0.0211

2.21%

PaiNN-S2EF-OC20-All

PaiNN

All

checkpoint | config | scale file

0.0294

0.91%

GemNet-OC-S2EF-OC20-2M

GemNet-OC

2M

checkpoint | config | scale file

0.0225

2.12%

GemNet-OC-S2EF-OC20-All

GemNet-OC

All

checkpoint | config | scale file

0.0179

4.56%

GemNet-OC-S2EF-OC20-All+MD

GemNet-OC

All+MD

checkpoint | config | scale file

0.0173

4.72%

GemNet-OC-Large-S2EF-OC20-All+MD

GemNet-OC-Large

All+MD

checkpoint | config | scale file

0.0164

5.34%

SCN-S2EF-OC20-2M

SCN

2M

checkpoint | config

0.0216

1.68%

SCN-t4-b2-S2EF-OC20-2M

SCN-t4-b2

2M

checkpoint | config

0.0193

2.68%

SCN-S2EF-OC20-All+MD

SCN

All+MD

checkpoint | config

0.0160

5.08%

eSCN-L4-M2-Lay12-S2EF-OC20-2M

eSCN-L4-M2-Lay12

2M

checkpoint | config

0.0191

2.55%

eSCN-L6-M2-Lay12-S2EF-OC20-2M

eSCN-L6-M2-Lay12

2M

checkpoint | config | exported

0.0186

2.66%

eSCN-L6-M2-Lay12-S2EF-OC20-All+MD

eSCN-L6-M2-Lay12

All+MD

checkpoint | config

0.0161

4.28%

eSCN-L6-M3-Lay20-S2EF-OC20-All+MD

eSCN-L6-M3-Lay20

All+MD

checkpoint | config

0.0139

6.64%

EquiformerV2-83M-S2EF-OC20-2M

EquiformerV2 (83M)

2M

checkpoint | config

0.0167

4.26%

EquiformerV2-31M-S2EF-OC20-All+MD

EquiformerV2 (31M)

All+MD

checkpoint | config

0.0142

6.20%

EquiformerV2-153M-S2EF-OC20-All+MD

EquiformerV2 (153M)

All+MD

checkpoint | config

0.0126

8.90%

S2EF models: optimized for force only#

Model Name

Model

Split

Download

val ID force MAE

SchNet-S2EF-force-only-OC20-All

SchNet

All

checkpoint

0.0443

DimeNet++-force-only-OC20-All

DimeNet++

All

checkpoint | config

0.0334

DimeNet++-Large-S2EF-force-only-OC20-All

DimeNet++-Large

All

checkpoint | config

0.02825

DimeNet++-S2EF-force-only-OC20-20M+Rattled

DimeNet++

20M+Rattled

checkpoint

0.0614

DimeNet++-S2EF-force-only-OC20-20M+MD

DimeNet++

20M+MD

checkpoint

0.0594

IS2RE models#

Model Name

Model

Split

Download

val ID energy MAE

CGCNN-IS2RE-OC20-10k

CGCNN

10k

checkpoint | config

0.9881

CGCNN-IS2RE-OC20-100k

CGCNN

100k

checkpoint | config

0.682

CGCNN-IS2RE-OC20-All

CGCNN

All

checkpoint | config

0.6199

DimeNet-IS2RE-OC20-10k

DimeNet

10k

checkpoint

1.0117

DimeNet-IS2RE-OC20-100k

DimeNet

100k

checkpoint

0.6658

DimeNet-IS2RE-OC20-all

DimeNet

All

checkpoint

0.5999

SchNet-IS2RE-OC20-10k

SchNet

10k

checkpoint | config

1.059

SchNet-IS2RE-OC20-100k

SchNet

100k

checkpoint | config

0.7137

SchNet-IS2RE-OC20-All

SchNet

All

checkpoint | config

0.6458

DimeNet++-IS2RE-OC20-10k

DimeNet++

10k

checkpoint | config

0.8837

DimeNet++-IS2RE-OC20-100k

DimeNet++

100k

checkpoint | config

0.6388

DimeNet++-IS2RE-OC20-All

DimeNet++

All

checkpoint | config

0.5639

PaiNN-IS2RE-OC20-All

PaiNN

All

checkpoint | config | scale file

0.5728

The Open Catalyst 2020 (OC20) dataset is licensed under a Creative Commons Attribution 4.0 License.

Please consider citing the following paper in any research manuscript using the OC20 dataset or pretrained models, as well as the original paper for each model:

@article{ocp_dataset,
    author = {Chanussot*, Lowik and Das*, Abhishek and Goyal*, Siddharth and Lavril*, Thibaut and Shuaibi*, Muhammed and Riviere, Morgane and Tran, Kevin and Heras-Domingo, Javier and Ho, Caleb and Hu, Weihua and Palizhati, Aini and Sriram, Anuroop and Wood, Brandon and Yoon, Junwoong and Parikh, Devi and Zitnick, C. Lawrence and Ulissi, Zachary},
    title = {Open Catalyst 2020 (OC20) Dataset and Community Challenges},
    journal = {ACS Catalysis},
    year = {2021},
    doi = {10.1021/acscatal.0c04525},
}

Open Catalyst 2022 (OC22)#

  • All configurations for these models are available in the configs/oc22 directory.

  • All of these models are trained on various splits of the OC22 S2EF / IS2RE datasets. For details, see https://arxiv.org/abs/2206.08917 and https://github.com/FAIR-Chem/fairchem/blob/main/DATASET.md.

  • All OC22 models released here are trained on DFT total energies, in contrast to the OC20 models listed above, which are trained on adsorption energies.

S2EF-Total models#

Model Name

Model

Training

Download

val ID force MAE

val ID energy MAE

GemNet-dT-S2EFS-OC22

GemNet-dT

OC22

checkpoint | config

0.032

1.127

GemNet-OC-S2EFS-OC22

GemNet-OC

OC22

checkpoint | config

0.030

0.563

GemNet-OC-S2EFS-OC20+OC22

GemNet-OC

OC20+OC22

checkpoint | config

0.027

0.483

GemNet-OC-S2EFS-nsn-OC20+OC22

GemNet-OC
(trained with enforce_max_neighbors_strictly=False, #467)

OC20+OC22

checkpoint | config

0.027

0.458

GemNet-OC-S2EFS-OC20->OC22

GemNet-OC

OC20->OC22

checkpoint | config

0.030

0.417

EquiformerV2-lE4-lF100-S2EFS-OC22

EquiformerV2 (\(\lambda_E\)=4, \(\lambda_F\)=100)

OC22

checkpoint | config

0.023

0.447

The Open Catalyst 2022 (OC22) dataset is licensed under a Creative Commons Attribution 4.0 License.

Please consider citing the following paper in any research manuscript using the OC22 dataset or pretrained models, as well as the original paper for each model:

@article{oc22_dataset,
    author = {Tran*, Richard and Lan*, Janice and Shuaibi*, Muhammed and Wood*, Brandon and Goyal*, Siddharth and Das, Abhishek and Heras-Domingo, Javier and Kolluru, Adeesh and Rizvi, Ammar and Shoghi, Nima and Sriram, Anuroop and Ulissi, Zachary and Zitnick, C. Lawrence},
    title = {The Open Catalyst 2022 (OC22) dataset and challenges for oxide electrocatalysts},
    journal = {ACS Catalysis},
    year={2023},
}

Open Direct Air Capture 2023 (ODAC23)#

  • All config files for the ODAC23 models are available in the configs/odac directory.

S2EF models#

Model Name

Model

Checkpoint

Config

SchNet-S2EF-ODAC

SchNet

checkpoint

config

DimeNet++-S2EF-ODAC

DimeNet++

checkpoint

config

PaiNN-S2EF-ODAC

PaiNN

checkpoint

config

GemNet-OC-S2EF-ODAC

GemNet-OC

checkpoint

config

eSCN-S2EF-ODAC

eSCN

checkpoint

config

EquiformerV2-S2EF-ODAC

EquiformerV2

checkpoint

config

EquiformerV2-Large-S2EF-ODAC

EquiformerV2 (Large)

checkpoint

config

IS2RE Direct models#

Model Name

Model

Checkpoint

Config

Gemnet-OC-IS2RE-ODAC

Gemnet-OC

checkpoint

config

eSCN-IS2RE-ODAC

eSCN

checkpoint

config

EquiformerV2-IS2RE-ODAC

EquiformerV2

checkpoint

config

The models in the table above were trained to predict relaxed energy directly. Relaxed energies can also be predicted by running structural relaxations using the S2EF models from the previous section.

IS2RS#

The IS2RS is solved by running structural relaxations using the S2EF models from the prior section.

The Open DAC 2023 (ODAC23) dataset is licensed under a Creative Commons Attribution 4.0 License.

Please consider citing the following paper in any research manuscript using the ODAC23 dataset:

@article{odac23_dataset,
    author = {Anuroop Sriram and Sihoon Choi and Xiaohan Yu and Logan M. Brabson and Abhishek Das and Zachary Ulissi and Matt Uyttendaele and Andrew J. Medford and David S. Sholl},
    title = {The Open DAC 2023 Dataset and Challenges for Sorbent Discovery in Direct Air Capture},
    year = {2023},
    journal={arXiv preprint arXiv:2311.00341},
}

Open Materials 2024 (OMat24)#

  • All config files for the OMat24 models are available in the configs/omat24 directory.

  • All models are equiformerV2 S2EFS models

Note in order to download any of the model checkpoints from the links below, you will need to first request access through the OMMAT24 Hugging Face page.

OMat pretrained models#

These checkpoints are trained on OMat24 only. Note that predictions are not Materials Project compatible.

Model Name

Checkpoint

Config

EquiformerV2-31M-OMat

checkpoint

config

EquiformerV2-86M-OMat

checkpoint

config

EquiformerV2-153M-OMat

checkpoint

config

MPTrj only models#

These models are trained only on the MPTrj dataset.

Model Name

Checkpoint

Config

EquiformerV2-31M-MP

checkpoint

config

EquiformerV2-31M-DeNS-MP

checkpoint

config

EquiformerV2-86M-DeNS-MP

checkpoint

config

EquiformerV2-153M-DeNS-MP

checkpoint

config

Finetuned OMat models#

These models are finetuned from the OMat pretrained checkpoints using MPTrj or MPTrj and sub-sampled trajectories from the 3D PBE Alexandria dataset, which we call Alex.

Model Name

Checkpoint

Config

EquiformerV2-31M-OMat-Alex-MP

checkpoint

config

EquiformerV2-86M-OMat-Alex-MP

checkpoint

config

EquiformerV2-153M-OMat-Alex-MP

checkpoint

config

Please consider citing the following work if you use OMat24 models in your work,

@article{barroso-luqueOpenMaterials20242024,
    title = {Open Materials 2024 (OMat24) Inorganic Materials Dataset and Models},
    author = {Barroso-Luque, Luis and Shuaibi, Muhammed and Fu, Xiang and Wood, Brandon M. and Dzamba, Misko and Gao, Meng and Rizvi, Ammar and Zitnick, C. Lawrence and Ulissi, Zachary W.},
    date = {2024-10-16},
    eprint = {2410.12771},
    eprinttype = {arXiv},
    doi = {10.48550/arXiv.2410.12771},
    url = {http://arxiv.org/abs/2410.12771},
}