List your company
ResearchFormerly WizWand

Image Classification on ImageNet 1k (pretraining)

81.19Top-1 AccuracyAdamCPR
75.9977.3478.6980.04Nov 15, 2023
Updated 8mo ago

Evaluation results

MethodLinks
AdamCPRBackbone=DeiT-Base, Number of Parameters=86M, Kappa WS (x lr-warmup)=1x2023.1181.19
AdamCPRBackbone=DeiT-Base, Number of Parameters=86M, Initialization=Kappa IP2023.1180.95
AdamWBackbone=DeiT-Base, Number of Parameters=86M, weight decay=0.52023.1180.56
AdamCPRBackbone=DeiT-Small, Number of Parameters=22M, Initialization=Kappa IP2023.1179.84
AdamCPRBackbone=DeiT-Small, Number of Parameters=22M, Kappa WS (x lr-warmup)=1x2023.1179.81
AdamCPRBackbone=DeiT-Base, Number of Parameters=86M, Kappa WS (x lr-warmup)=2x2023.1179.61
AdamCPRBackbone=DeiT-Small, Number of Parameters=22M, Kappa WS (x lr-warmup)=2x2023.1179.33
AdamWBackbone=DeiT-Small, Number of Parameters=22M, weight decay=0.52023.1179.16
AdamWBackbone=DeiT-Small, Number of Parameters=22M, weight decay=0.052023.1179.03
AdamWBackbone=DeiT-Base, Number of Parameters=86M, weight decay=0.052023.1178.59
AdamCPRBackbone=DeiT-Small, Number of Parameters=22M, Kappa WS (x lr-warmup)=4x2023.1178.04
AdamWBackbone=DeiT-Small, Number of Parameters=22M, weight decay=0.0052023.1176.97
AdamWBackbone=DeiT-Base, Number of Parameters=86M, weight decay=0.0052023.1176.19