results on cifar-100

[chenchen9407]

I tested it with cifar100. However, it can not achieve the desired effect, especially in the incremental stage.
train 140 epoch：
task id = 5 @ACC1 = 25.73000, acc1s = [77.32000010986329, 27.400000030517578, 32.328571446010045, 29.5625, 30.6, 25.73]