[問題] class用法

[Ohwil (Ohwil)]

https://www.pythonmorsels.com/p/22cjj/

各位好, 想問上面line 345 vs line 346  一個true 一個False

都使用forward但是output不同....

貼上來如下

    den = Dense(784, 150)
    den.weight = model.layers[0].params[0]
    den.bias = model.layers[0].params[1]
->  oout1_0 = den.forward(out1[0])

    denn = NN()
    denn.add_layer(Dense(784, 150))
    denn.params.append(model.layers[0].params)
->  oout1, _ = denn.forward(out1[0])
    oout3    = np.dot(out1[0], model.layers[0].params[0]) +
model.layers[0].params[1]

    print((oout1 == oout1_0).all())  # False
    print((oout3 == oout1_0).all())  # True

而
class Dense():
    def __init__(self, input_node, output_node):
        ......
        ......

    def forward(self, x):
        self.x = x   # self.x is the input
        self.output = np.dot(x, self.weight) + self.bias # generate output
from self.x
        return self.output


另外想問 save_parameter跟load parameter有沒有寫錯呢orz


--
※ 發信站: 批踢踢實業坊(ptt.cc), 來自: 114.140.90.201 (臺灣)
※ 文章網址: https://www.ptt.cc/bbs/Python/M.1667400516.A.C0B.html
推 lycantrope: 因為denn.append_layer的Dense，沒有給一樣的weight