默认初始化效果（效果极差

自己抄的初始化(效果一般）

`class relu(fluid.dygraph.Layer):
definit(self):
super(relu,self).init()
def forward(self, x):
prelu = fluid.dygraph.PRelu("all")
x = prelu(x)
return x

class upsample(fluid.dygraph.Layer):
definit(self, channel=64, scale = 2):
super(upsample, self).init()
self.scale = scale
self.n1 = Conv2D(channel, channelself.scaleself.scale, 3,padding=1)
def forward(self, x):
x = self.n1(x)
x = fluid.layers.pixel_shuffle(x,upscale_factor=self.scale)
return x

class MSRB(fluid.dygraph.Layer):
definit(self,channels=64):
super(MSRB, self).init()
self.conv_3_1 = Conv2D (channels,channels,3,padding = 1,act='relu')
self.conv_3_2 = Conv2D (channels2,channels2,3,padding = 1,act='relu')
self.conv_5_1 = Conv2D (channels,channels,5,padding = 2,act='relu')
self.conv_5_2 = Conv2D (channels2,channels2,5,padding = 2,act='relu')
self.confusion = Conv2D (channels*4,channels,1,padding = 0)
def forward(self, x):
input_1 = x
output_3_1 = self.conv_3_1(input_1)
output_5_1 = self.conv_5_1(input_1)
input_2 = fluid.layers.concat([output_3_1, output_5_1], axis=1)
output_3_2 = self.conv_3_2(input_2)
output_5_2 = self.conv_5_2(input_2)
input_3 = fluid.layers.concat([output_3_2, output_5_2], axis=1)
output = self.confusion(input_3)
output += x
return output
class MSRN(fluid.dygraph.Layer):
definit(self):
super(MSRN, self).init()
self.n_blocks = 1

# define head module

    modules_head = [Conv2D(3, 64, 3,padding = 1)]

    # define body module
    modules_body = paddle.fluid.dygraph.LayerList([MSRB()for i in range(self.n_blocks)])

    # define tail module
    modules_tail = [
        Conv2D(64 * (self.n_blocks + 1), 64, 1),
        Conv2D(64, 64, 3,padding = 1),
        upsample(),
        Conv2D(64, 3, 3,padding = 1)]
    self.head = fluid.dygraph.Sequential(*modules_head)
    self.body = fluid.dygraph.Sequential(*modules_body)
    self.tail = fluid.dygraph.Sequential(*modules_tail)

def forward(self, x):
    x = self.head(x)
    res = x

    MSRB_out = []
    for i in range(self.n_blocks):
        x = self.body[i](x)
        MSRB_out.append(x)
    MSRB_out.append(res)

    res = fluid.layers.concat(MSRB_out,axis=1)
    x = self.tail(res)
    return x  `