v1.0.0 fine tune result

# model definition
class Resnet101(nn.Module):
    def __init__(self):
        super(Resnet101, self).__init__()

        self.resnet = torchvision.models.resnet101(
            weights=torchvision.models.ResNet101_Weights.IMAGENET1K_V2,
        )
        # for p in self.resnet.parameters():
        #    p.requires_grad = False
        self.resnet = nn.Sequential(
            *list(self.resnet.children())[:-1],
            nn.Flatten(),
            nn.GELU(),
            nn.Linear(2048, 2048),
        )


    def forward(self, x):
        return self.resnet(x)class Resnet101(nn.Module):
    def __init__(self):
        super(Resnet101, self).__init__()

        self.resnet = torchvision.models.resnet101(
            weights=torchvision.models.ResNet101_Weights.IMAGENET1K_V2,
        )
        for p in self.resnet.parameters():
            p.requires_grad = False
        self.resnet = nn.Sequential(
            *list(self.resnet.children())[:-1],
            nn.Flatten(),
            nn.GELU(),
            nn.Linear(2048, 2048),
        )


    def forward(self, x):
        return self.resnet(x)