IGCV3-pytorch
IGCV3-pytorch copied to clipboard
About the inverted residual block
Hello, I found the original paper of the inverted residual block is "G1x1_Cw3x3_ReLU_G1x1", but I found the implementation of your code is "G1x1_ReLU_Cw3x3_ReLU_G1x1". And the paper said "G1x1_ReLU_Cw3x3_ReLU_G1x1" would hurt the performance of the network. Is there something wrong?
do you get the same result on CIFAR100 as mentioned in the paper