NASViT: Neural Architecture Search for Efficient Vision Transformers with Gradient Conflict aware Supernet Training

Open paper
Benchmark Model Rank Results
image-classification-on-imagenetNASViT (supernet)#472Top 1 Accuracy: 82.9%GFLOPs: 1.881
image-classification-on-imagenetNASViT-A5#584Top 1 Accuracy: 81.8%GFLOPs: 0.757
image-classification-on-imagenetNASViT-A4#618Top 1 Accuracy: 81.4%GFLOPs: 0.591
image-classification-on-imagenetNASViT-A3#643Top 1 Accuracy: 81.0%GFLOPs: 0.528
image-classification-on-imagenetNASViT-A2#668Top 1 Accuracy: 80.5%GFLOPs: 0.421
image-classification-on-imagenetNASViT-A1#705Top 1 Accuracy: 79.7%GFLOPs: 0.309
image-classification-on-imagenetNASViT-A0#797Top 1 Accuracy: 78.2%GFLOPs: 0.208
neural-architecture-search-on-imagenetNASViT-A5#7Top-1 Error Rate: 18.2Accuracy: 81.8FLOPs: 757M
neural-architecture-search-on-imagenetNASViT-A4#8Top-1 Error Rate: 18.6Accuracy: 81.4FLOPs: 591M
neural-architecture-search-on-imagenetNASViT-A3#9Top-1 Error Rate: 19.0Accuracy: 81.0FLOPs: 528M
neural-architecture-search-on-imagenetNASViT-A2#14Top-1 Error Rate: 19.5Accuracy: 80.5FLOPs: 421M
neural-architecture-search-on-imagenetNASViT-A1#25Top-1 Error Rate: 20.3Accuracy: 79.7FLOPs: 309M
neural-architecture-search-on-imagenetNASViT-A0#47Top-1 Error Rate: 21.8Accuracy: 78.2FLOPs: 208M

Rank counts only results with a code link.