Image Classification on ImageNet 1k (val) (Accuracy & Relative FLOPs)
79.3AccuracyBLAST_3
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| BLAST_3Backbone=ViT-Base, Image Resolution=224x224, Patch Size=16x16, Training=From scratch, Blocks=3x32024.10 | 79.3 | 27.8 | |
| Low-RankBackbone=ViT-Base, Image Resolution=224x224, Patch Size=16x16, Training=From scratch2024.10 | 78.9 | 33.5 | |
| MonarchBackbone=ViT-Base, Image Resolution=224x224, Patch Size=16x16, Training=From scratch2024.10 | 78.9 | 33.5 | |
| Dense ViT-BaseBackbone=ViT-Base, Image Resolution=224x224, Patch Size=16x16, Training=From scratch2024.10 | 78.7 | 100 | |
| Gaudi-GBLRBackbone=ViT-Base, Image Resolution=224x224, Patch Size=16x16, Training=From scratch2024.10 | 78.5 | 32.8 |