List your company
ResearchFormerly WizWand

Structure to Energy and Forces (S2EF) on OC20 (val average)

0.297Force MAE (meV/Å)SchNet
-2.8231218.2376939.298560.35931Apr 6, 2022Jun 18, 2022Aug 31, 2022Nov 12, 2022Jan 25, 2023Apr 8, 2023Jun 21, 2023
Updated 8mo ago

Evaluation results

MethodLinks
SchNetTrain set=OC20, Throughput (Samples / GPU sec)=5492022.040.2975680.06—
EquiformerV2Training set=OC20 S2EF-All+MD, Number of parameters=153M, Training time (GPU-days)=1368, Throughput (Samples/GPU sec.)=1.8, lambda_E=22023.0614.6230——
EquiformerV2Training set=OC20 S2EF-All+MD, Number of parameters=153M, Training time (GPU-days)=1571, Throughput (Samples/GPU sec.)=1.8, lambda_E=42023.0615227——
EquiformerV2Training set=OC20 S2EF-All, Number of parameters=153M, Training time (GPU-days)=1368, Throughput (Samples/GPU sec.)=1.8, lambda_E=22023.0615.7236——
EquiformerV2Training set=OC20 S2EF-All+MD, Number of parameters=31M, Training time (GPU-days)=705, Throughput (Samples/GPU sec.)=7.1, lambda_E=42023.0616.3232——
eSCN L=6 K=20Training set=OC20 S2EF-All, Number of parameters=200M, Training time (GPU-days)=600, Throughput (Samples/GPU sec.)=2.92023.0617.1242——
eSCN L=6 K=20Training set=OC20 S2EF-All+MD, Number of parameters=200M, Training time (GPU-days)=568, Throughput (Samples/GPU sec.)=2.92023.0617.1243——
SCN L=8 K=20Training set=OC20 S2EF-All+MD, Number of parameters=271M, Training time (GPU-days)=12802023.0617.2237——
SCN L=8 K=20Training set=OC20 S2EF-All, Number of parameters=271M, Training time (GPU-days)=6452023.0617.7244——
SCN L=6 K=16 (4-tap 2-band)Training set=OC20 S2EF-All+MD, Number of parameters=168M, Training time (GPU-days)=4142023.0617.8228——
GemNet-OC-L-FTrain set=OC20 + OC-MD, Throughput (Samples / GPU sec)=3.22022.04202520.6872.51
GemNet-OC-L-FTraining set=OC20 S2EF-All+MD, Number of parameters=216M, Training time (GPU-days)=765, Throughput (Samples/GPU sec.)=3.22023.0620252——
GemNet-OCTrain set=OC20, Throughput (Samples / GPU sec)=18.32022.0421.72440.6622.07
GemNet-OCTraining set=OC20 S2EF-All, Number of parameters=39M, Training time (GPU-days)=336, Throughput (Samples/GPU sec.)=18.32023.0621.7244——
GemNet-OC-L-ETrain set=OC20 + OC-MD, Throughput (Samples / GPU sec)=7.52022.0422.12390.6622.3
GemNet-OC-L-ETraining set=OC20 S2EF-All+MD, Number of parameters=56M, Training time (GPU-days)=640, Throughput (Samples/GPU sec.)=7.52023.0622.1239——
GemNet-OCTrain set=OC-2M2022.0425.72860.5981.06
GemNet-dTTrain set=OC20, Throughput (Samples / GPU sec)=25.82022.0427.23150.5940.54
GemNet-dTTraining set=OC20 S2EF-All, Number of parameters=32M, Training time (GPU-days)=492, Throughput (Samples/GPU sec.)=25.82023.0627.2315——
GemNet-dTTrain set=OC-2M2022.0429.53580.5570.61
DimeNet++-L-F+ETrain set=OC20, Throughput (Samples / GPU sec)=4.62022.0432.85150.5410
DimeNet++-L-F+ETraining set=OC20 S2EF-All, Number of parameters=10.7M, Training time (GPU-days)=1600, Throughput (Samples/GPU sec.)=4.62023.0632.8515——
ForceNet-largeTrain set=OC20, Throughput (Samples / GPU sec)=15.32022.0433.5—0.515—
SpinConvTrain set=OC-2M2022.0436.24060.4790.13
SpinConvTrain set=OC20, Throughput (Samples / GPU sec)=62022.0441.23710.4730.05
SpinConvTraining set=OC20 S2EF-All, Number of parameters=8.5M, Training time (GPU-days)=275, Throughput (Samples/GPU sec.)=62023.0641.2371——
SchNetTraining set=OC20 S2EF-All, Number of parameters=9.1M, Training time (GPU-days)=194, Throughput (Samples/GPU sec.)=5492023.0656.8549——
DimeNet++Train set=OC-2M2022.0465.78050.2170.01
CGCNNTrain set=OC202022.04745900.1420.01
SchNetTrain set=OC-2M2022.0478.31,4000.1090