Image Classification on TinyImageNet (test) with Memory Efficiency
64.23AccuracyTRGL
Evaluation results
| Method | Links | ||
|---|---|---|---|
| TRGLArchitecture=ResNet-152, Number of modules=4, Training mode=Parallel module-wise2023.09 | 64.23 | 10 | |
| SedonaArchitecture=ResNet-152, Number of modules=4, Training mode=Parallel module-wise2023.09 | 64.1 | — | |
| VanGLArchitecture=ResNet-152, Number of modules=4, Training mode=Parallel module-wise2023.09 | 63.87 | 21 | |
| TRGLArchitecture=ResNet-101, Number of modules=4, Training mode=Parallel module-wise2023.09 | 63.71 | 11 | |
| VanGLArchitecture=ResNet-101, Number of modules=4, Training mode=Parallel module-wise2023.09 | 63.64 | 24 | |
| E2EArchitecture=ResNet-152, Training mode=End-to-end2023.09 | 62.32 | — | |
| E2EArchitecture=ResNet-101, Training mode=End-to-end2023.09 | 62.01 | — | |
| TRGLArchitecture=ResNet-50, Number of modules=4, Training mode=Parallel module-wise2023.09 | 60.3 | 20 | |
| SedonaArchitecture=ResNet-101, Number of modules=4, Training mode=Parallel module-wise2023.09 | 59.12 | — | |
| E2EArchitecture=VGG-19, Training mode=End-to-end2023.09 | 58.74 | — | |
| VanGLArchitecture=ResNet-50, Number of modules=4, Training mode=Parallel module-wise2023.09 | 58.43 | 26 | |
| E2EArchitecture=ResNet-50, Training mode=End-to-end2023.09 | 58.1 | — | |
| DGLArchitecture=ResNet-152, Number of modules=4, Training mode=Parallel module-wise2023.09 | 57.64 | — | |
| TRGLArchitecture=VGG-19, Number of modules=4, Training mode=Parallel module-wise2023.09 | 57.28 | 21 | |
| SedonaArchitecture=VGG-19, Number of modules=4, Training mode=Parallel module-wise2023.09 | 56.56 | — | |
| VanGLArchitecture=VGG-19, Number of modules=4, Training mode=Parallel module-wise2023.09 | 56.17 | 27 | |
| SedonaArchitecture=ResNet-50, Number of modules=4, Training mode=Parallel module-wise2023.09 | 54.4 | — | |
| DGLArchitecture=ResNet-50, Number of modules=4, Training mode=Parallel module-wise2023.09 | 53.96 | — | |
| PredSimArchitecture=ResNet-101, Number of modules=4, Training mode=Parallel module-wise2023.09 | 53.92 | — | |
| DGLArchitecture=ResNet-101, Number of modules=4, Training mode=Parallel module-wise2023.09 | 53.8 | — | |
| PredSimArchitecture=ResNet-152, Number of modules=4, Training mode=Parallel module-wise2023.09 | 51.76 | — | |
| DGLArchitecture=VGG-19, Number of modules=4, Training mode=Parallel module-wise2023.09 | 51.4 | — | |
| PredSimArchitecture=ResNet-50, Number of modules=4, Training mode=Parallel module-wise2023.09 | 47.48 | — | |
| PredSimArchitecture=VGG-19, Number of modules=4, Training mode=Parallel module-wise2023.09 | 44.7 | — |