model.parameters()的理解与使用

model.parameters()保存的是Weights和Bais参数的值。
首先定义一个模型

#design Model
class NeuralNetwork(nn.Module):
    def __init__(self):
        super(NeuralNetwork,self).__init__()
        self.flatten = nn.Flatten()
        self.linear_relu_stack = nn.Sequential(
            nn.Linear(2*2, 5),
            nn.ReLU(),
            nn.Linear(5,3)
            nn.ReLU()
        )
    
    def forward(self, x):
        x = self.flatten(x)
        logits = self.linear_relu_stack(x)
        return logits
    
model = NeuralNetwork().to(device)
print(model)

'''
#print：
NeuralNetwork(
  (flatten): Flatten()
  (linear_relu_stack): Sequential(
    (0): Linear(in_features=4, out_features=5, bias=True)
    (1): ReLU()
    (2): Linear(in_features=5, out_features=3, bias=True)
    (3): ReLU()
  )
)
'''

输入一个2 * 2的input（比如2*2的图片），构建长度为5的隐藏全连接层，输出3类（预测图片的label）。
$$H=W_{i}X+b;O=\sigma (W_{o}H+b)$$

查看model.parameters()的参数

print(list(model.parameters()))
'''
[Parameter containing:                        #输入层的 W，每列即为对应x乘的w（维度为5）；Wi = [w1,w2,w3,w4]
tensor([[-0.2857,  0.2448,  0.1023, -0.4781],      
        [-0.2275,  0.3597,  0.3495,  0.0163],
        [-0.4321, -0.2033,  0.1903, -0.1599],
        [ 0.4215,  0.3982,  0.0478, -0.4342],
        [-0.3891,  0.2578,  0.1874, -0.1962]], device='cuda:0',
       requires_grad=True), Parameter containing:
tensor([-0.2413, -0.1015, -0.4152,  0.2088,  0.4680], device='cuda:0', 
       requires_grad=True), Parameter containing:			#隐藏层的 W,每列为h对应乘的w(维度为3); Wo = [w1,w2,w3,w4,w5] 
tensor([[-3.5943e-01, -1.1033e-01,  8.3123e-02,  7.8943e-02, -3.6048e-01],
        [ 1.3593e-01,  4.2951e-01, -3.7831e-01,  2.6139e-01,  1.0279e-01],
        [ 2.5864e-01,  4.1339e-01,  2.4220e-04, -3.3391e-01,  7.3478e-02]],
       device='cuda:0', requires_grad=True), Parameter containing:
tensor([0.3463, 0.0749, 0.0696], device='cuda:0', requires_grad=True)]
'''

实际上parameters里存的就是weight，parameters()会返回一个生成器（迭代器）