Initial Structure to Relaxed Structure (IS2RS) on Open Catalyst OC20 (test)
0.117AFbTPaiNN
Evaluation results
| Method | Links | |||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| PaiNNTrain set=OC20, Throughput (Samples / GPU sec)=602022.04 | 0.117 | 0.485 | — | — | — | — | — | — | — | — | — | — | — | |
| ForceNet-largeTrain set=OC20, Throughput (Samples / GPU sec)=15.32022.04 | 0.127 | 0.496 | — | — | — | — | — | — | — | — | — | — | — | |
| SchNetTrain set=OC20, Throughput (Samples / GPU sec)=5492022.04 | 0.144 | — | — | — | — | — | — | — | — | — | — | — | — | |
| SchNetTraining set=OC20 S2EF-All, Number of parameters=9.1M2023.06 | 0.144 | 0.749 | — | — | — | — | — | — | — | — | — | — | — | |
| GemNet-dTTrain set=OC-2M2022.04 | 0.167 | 0.548 | — | — | — | — | — | — | — | — | — | — | — | |
| SpinConvTrain set=OC20, Throughput (Samples / GPU sec)=62022.04 | 0.167 | 0.536 | — | — | — | — | — | — | — | — | — | — | — | |
| SpinConvTraining set=OC20 S2EF-All, Number of parameters=8.5M2023.06 | 0.167 | 0.536 | — | — | — | — | — | — | — | — | — | — | — | |
| GemNet-OCTrain set=OC-2M2022.04 | 0.196 | 0.564 | — | — | — | — | — | — | — | — | — | — | — | |
| DimeNet++-L-F+ETrain set=OC20, Throughput (Samples / GPU sec)=4.62022.04 | 0.217 | 0.517 | — | — | — | — | — | — | — | — | — | — | — | |
| DimeNet++-L-F+ETraining set=OC20 S2EF-All, Number of parameters=10.7M2023.06 | 0.217 | 0.517 | — | — | — | — | — | — | — | — | — | — | — | |
| GemNet-dTTrain set=OC20, Throughput (Samples / GPU sec)=25.82022.04 | 0.276 | 0.587 | — | — | — | — | — | — | — | — | — | — | — | |
| GemNet-dTTraining set=OC20 S2EF-All, Number of parameters=32M2023.06 | 0.276 | 0.587 | — | — | — | — | — | — | — | — | — | — | — | |
| GemNet-XLTrain set=OC20, Throughput (Samples / GPU sec)=1.52022.04 | 0.308 | 0.627 | — | — | — | — | — | — | — | — | — | — | — | |
| GemNet-OCTrain set=OC20, Throughput (Samples / GPU sec)=18.32022.04 | 0.353 | 0.603 | — | — | — | — | — | — | — | — | — | — | — | |
| GemNet-OCTraining set=OC20 S2EF-All, Number of parameters=39M2023.06 | 0.353 | 0.603 | — | — | — | — | — | — | — | — | — | — | — | |
| SCN L=8 K=20Training set=OC20 S2EF-All, Number of parameters=271M2023.06 | 0.403 | 0.671 | — | — | — | — | — | — | — | — | — | — | — | |
| GemNet-OC-L-FTrain set=OC20 + OC-MD, Throughput (Samples / GPU sec)=3.22022.04 | 0.406 | 0.604 | — | — | — | — | — | — | — | — | — | — | — | |
| GemNet-OC-L-FTraining set=OC20 S2EF-All+MD, Number of parameters=216M2023.06 | 0.406 | 0.604 | — | — | — | — | — | — | — | — | — | — | — | |
| SCN L=6 K=16 (4-tap 2-band)Training set=OC20 S2EF-All+MD, Number of parameters=168M2023.06 | 0.433 | 0.649 | — | — | — | — | — | — | — | — | — | — | — | |
| SCN L=8 K=20Training set=OC20 S2EF-All+MD, Number of parameters=271M2023.06 | 0.436 | 0.675 | — | — | — | — | — | — | — | — | — | — | — | |
| EquiformerV2Training set=OC20 S2EF-All+MD, Number of parameters=31M, lambda_E=42023.06 | 0.476 | 0.683 | — | — | — | — | — | — | — | — | — | — | — | |
| eSCN L=6 K=20Training set=OC20 S2EF-All, Number of parameters=200M2023.06 | 0.485 | 0.657 | — | — | — | — | — | — | — | — | — | — | — | |
| eSCN L=6 K=20Training set=OC20 S2EF-All+MD, Number of parameters=200M2023.06 | 0.503 | 0.667 | — | — | — | — | — | — | — | — | — | — | — | |
| EquiformerV2Training set=OC20 S2EF-All, Number of parameters=153M, lambda_E=22023.06 | 0.53 | 0.69 | — | — | — | — | — | — | — | — | — | — | — | |
| EquiformerV2Training set=OC20 S2EF-All+MD, Number of parameters=153M, lambda_E=42023.06 | 0.544 | 0.694 | — | — | — | — | — | — | — | — | — | — | — | |
| EquiformerV2Training set=OC20 S2EF-All+MD, Number of parameters=153M, lambda_E=22023.06 | 0.554 | 0.698 | — | — | — | — | — | — | — | — | — | — | — | |
| SpinConv#Params=8.9M, Training Dataset=S2EF-All2022.03 | 16.67 | 53.62 | 0.05 | — | — | — | — | — | — | — | — | — | — | |
| DimeNet++#Params=1.8M, Training Dataset=S2EF 20M + MD2022.03 | 17.15 | 47.72 | 0.15 | — | — | — | — | — | — | — | — | — | — | |
| DimeNet++-large#Params=10.8M, Training Dataset=S2EF-All2022.03 | 21.82 | 51.68 | 0.4 | — | — | — | — | — | — | — | — | — | — | |
| GemNet-T#Params=31M, Training Dataset=S2EF-All2022.03 | 27.6 | 58.68 | 0.7 | — | — | — | — | — | — | — | — | — | — | |
| GemNet-XL#Params=300M, Training Dataset=S2EF-All2022.03 | 30.82 | 62.65 | 0.9 | — | — | — | — | — | — | — | — | — | — | |
| DimeNet++-XL#Params=240M, Training Dataset=S2EF 20M + MD2022.03 | 33.44 | 59.21 | 1.25 | — | — | — | — | — | — | — | — | — | — | |
| DimeNet++Inference time=407.6h2021.06 | — | — | — | 17.52 | 14.67 | 14.32 | 14.43 | 15.23 | 48.76 | 45.19 | 48.59 | 53.14 | 48.92 | |
| DimeNet++-largeInference time=814.6h2021.06 | — | — | — | 25.65 | 20.73 | 20.24 | 20.67 | 21.82 | 52.45 | 48.47 | 50.99 | 54.82 | 51.68 | |
| ForceNetInference time=75.1h2021.06 | — | — | — | 10.75 | 7.74 | 7.54 | 7.78 | 8.45 | 46.83 | 41.26 | 46.45 | 49.6 | 46.04 | |
| ForceNet-largeInference time=186.9h2021.06 | — | — | — | 14.77 | 12.23 | 12.16 | 11.46 | 12.66 | 50.59 | 45.16 | 49.8 | 52.94 | 49.62 | |
| SchNetInference time=54.1h2021.06 | — | — | — | 5.28 | 2.82 | 2.62 | 2.73 | 3.36 | 32.49 | 28.59 | 30.99 | 35.08 | 31.79 | |
| SpinConv (force-centric)Inference time=263.2h2021.06 | — | — | — | 21.1 | 15.7 | 15.86 | 14.01 | 16.67 | 53.68 | 48.87 | 53.92 | 58.03 | 53.62 |