Training Efficiency on ResNet architecture
22.06Latency (s)FedAvg
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| FedAvgBackbone=ResNet, Iterations per round=100, Hardware=NVIDIA GeForce RTX 2080 Ti, Platform=Pytorch 2.0.1, Cuda version=11.72024.09 | 22.06 | 1 | |
| FedDynBackbone=ResNet, Iterations per round=100, Hardware=NVIDIA GeForce RTX 2080 Ti, Platform=Pytorch 2.0.1, Cuda version=11.72024.09 | 34.03 | 1.54 | |
| SCAFFOLDBackbone=ResNet, Iterations per round=100, Hardware=NVIDIA GeForce RTX 2080 Ti, Platform=Pytorch 2.0.1, Cuda version=11.72024.09 | 36.01 | 1.63 | |
| A-FedPDBackbone=ResNet, Iterations per round=100, Hardware=NVIDIA GeForce RTX 2080 Ti, Platform=Pytorch 2.0.1, Cuda version=11.72024.09 | 37.61 | 1.7 | |
| FedSAMBackbone=ResNet, Iterations per round=100, Hardware=NVIDIA GeForce RTX 2080 Ti, Platform=Pytorch 2.0.1, Cuda version=11.72024.09 | 39.21 | 1.77 | |
| FedSpeedBackbone=ResNet, Iterations per round=100, Hardware=NVIDIA GeForce RTX 2080 Ti, Platform=Pytorch 2.0.1, Cuda version=11.72024.09 | 56.1 | 2.54 | |
| A-FedPDSAMBackbone=ResNet, Iterations per round=100, Hardware=NVIDIA GeForce RTX 2080 Ti, Platform=Pytorch 2.0.1, Cuda version=11.72024.09 | 60.63 | 2.74 |