反向传播练习出现神奇一幕

Question

软件：Jupyter notebook

反向传播练习，以下代码

import numpy as np

# 编写乘法层类
class MulLayer:

    def __init__(self):
        # 编写需要初始化的变量
        self.x = None
        self.y = None

    def forward(self, x, y):
        # 编写前向传播过程
        self.x = x
        self.y = y
        out = x * y
        return out

    def backward(self, dout):
        # 编写反向传播过程
        dx = dout *self.y
        dy = dout *self.x
        return dx, dy

# 编写加法层类   
class AddLayer:
    def __init__(self):
        # 编写需要初始化的变量
        pass

    def forward(self, x, y):
        # 编写前向传播过程
        out = x + y
        return out

    def backward(self, dout):
        # 编写反向传播过程
        dx = dout * 1
        dy = dout * 1
        return dx, dy
    
# 前向传播的输入变量
apple = 100
apple_num = 2
orange = 150
orange_num = 3
tax = 1.1

# layer功能层实例化
mul_apple_layer = MulLayer()
mul_orange_layer = MulLayer()
add_apple_orange_layer = AddLayer()
mul_tax_layer = MulLayer()

# forward：写出前向消费金额计算过程
apple_price = mul_apple_layer.forward(apple, apple_num)  # (1)
orange_price = mul_orange_layer.forward(orange, orange_num)  # (2)
all_price = add_apple_orange_layer.forward(apple_price, orange_price)  # (3)
price = mul_tax_layer.forward(all_price, tax)  # (4)

# backward：写出反向传播求导的过程
dprice = 1
dall_price, dtax = mul_tax_layer.backward(dprice)  # (4)
dapple_price, dorange_price = add_apple_orange_layer.backward(dall_price)  # (3)
dorange, dorange_num = mul_orange_layer.backward(dorange_price)  # (2)
dapple, dapple_num = mul_apple_layer.backward(dapple_price)  # (1)

print("price:", int(price))
print("dApple:", dapple)
print("dApple_num:", int(dapple_num))
print("dOrange:", dorange)
print("dOrange_num:", int(dorange_num))
print("dTax:", dtax)

合在一个编译框里，就可以输出，结果如下

price: 715
dApple: 2.2
dApple_num: 110
dOrange: 3.3000000000000003
dOrange_num: 165
dTax: 650

从apple行之前，分开成两个编译框，分别运行，就会报错如下

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-5-30ce15d60af2> in <module>()
     20 # backward：写出反向传播求导的过程
     21 dprice = 1
---> 22 dall_price, dtax = mul_tax_layer.backward(dprice)  # (4)
     23 dapple_price, dorange_price = add_apple_orange_layer.backward(dall_price)  # (3)
     24 dorange, dorange_num = mul_orange_layer.backward(dorange_price)  # (2)

<ipython-input-4-f8bcb7ea494e> in backward(self, dout)
     18     def backward(self, dout):
     19         # 编写反向传播过程
---> 20         dx = dout *self.y
     21         dy = dout *self.x
     22         return dx, dy

TypeError: unsupported operand type(s) for *: 'int' and 'NoneType'

一模一样的代码，复制粘贴的，有盆友懂吗？为什么会出现这样的问题？

Answer 1

合在一起的时候，通过forward()方法给实例的属性 x 和 y 赋了值，所以backward()方法才能进行计算
分开以后，实例化没有给对象的x和y属性赋值，所以self.x和self.y默认都是None，所以无法计算