Peak GPU Memory (GB) on Class-Incremental Learning on CIFAR100
0.66Peak GPU Memory (GB)APER
Evaluation results
| Method | Links | |
|---|---|---|
| APEROptimization Protocol=ZO Optimization, Batch size=482025.10 | 0.66 | |
| Learnable Cls.Optimization Protocol=ZO Optimization, Batch size=482025.10 | 0.69 | |
| Adapter + Learnable Cls.Optimization Protocol=ZO Optimization, Batch size=482025.10 | 0.69 | |
| LAEOptimization Protocol=ZO Optimization, Batch size=482025.10 | 0.69 | |
| Learnable Cls.Optimization Protocol=FO Optimization, Batch size=482025.10 | 0.7 | |
| ZO-FCOptimization Protocol=ZO Optimization, Batch size=482025.10 | 0.7 | |
| EASEOptimization Protocol=ZO Optimization, Batch size=482025.10 | 0.87 | |
| InfLoRAOptimization Protocol=ZO Optimization, Batch size=482025.10 | 0.88 | |
| APEROptimization Protocol=FO Optimization, Batch size=482025.10 | 4.25 | |
| Adapter + Learnable Cls.Optimization Protocol=FO Optimization, Batch size=482025.10 | 4.35 | |
| EASEOptimization Protocol=FO Optimization, Batch size=482025.10 | 4.38 | |
| LAEOptimization Protocol=FO Optimization, Batch size=482025.10 | 4.48 | |
| InfLoRAOptimization Protocol=FO Optimization, Batch size=482025.10 | 4.98 | |
| L2POptimization Protocol=FO Optimization, Batch size=482025.10 | 5.58 | |
| CODA-PromptOptimization Protocol=FO Optimization, Batch size=482025.10 | 7.21 |