Supernet Training on CIFAR-10 NAS-Bench-201 (test)
94.37Top-1 AccuracyOptimal
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Optimal2025.03 | 94.37 | — | — | — | |
| FSNAS + OursTraining Mode=with CaLR and MS2025.03 | 93.63 | 0.767 | 1,925 | 16.12 | |
| FairNAS + OursTraining Mode=with CaLR and MS2025.03 | 93.52 | 0.828 | 1,903 | 5.15 | |
| SPOS + OursTraining Mode=with CaLR and MS2025.03 | 93.5 | 0.814 | 1,901 | 2.69 | |
| FSNAS2025.03 | 93.43 | 0.729 | 1,901 | 16.04 | |
| SPOS2025.03 | 93.12 | 0.751 | 1,875 | 2.67 | |
| FairNAS2025.03 | 92.13 | 0.766 | 1,877 | 5.13 |