pytorch train mode

Prequel judgment:

    def forward(self, x): # [3,112,112]

        if self.training:
            print("train")
        else:
            print("eval")

transfer:

if __name__ == "__main__":
    net = MFN_85m()

    net.eval()