Created by: ruyijidan

下面两种方式，训练时现存占用都差不多，但是第一种方式保存的模型要比第二种方式小很多，这是为什么呢？

mylayers1 = []
for _ in range(3):
    layers = fluid.dygraph.Sequential(Linear(100, 100),
                                      (Linear(100, 100)),
                                      (Linear(100, 100)),
                                      (Linear(100, 10)))
    mylayers.append(layers)

for i in range(3):
    x = mylayers[i](x)

#################################################################
mylayers2_0 = fluid.dygraph.Sequential(Linear(100, 100),
                                       (Linear(100, 100)),
                                       (Linear(100, 100)),
                                       (Linear(100, 10)))
mylayers2_1 = fluid.dygraph.Sequential(Linear(100, 100),
                                       (Linear(100, 100)),
                                       (Linear(100, 100)),
                                       (Linear(100, 10)))
mylayers2_2 = fluid.dygraph.Sequential(Linear(100, 100),
                                       (Linear(100, 100)),
                                       (Linear(100, 100)),
                                       (Linear(100, 10)))

x = mylayers2_0(x)
x = mylayers2_1(x)
x = mylayers2_2(x)