LeNet-1D-V模型代码分析及错误修正

本文分析了LeNet-1D-V模型代码，并指出其在forward方法中缺少target输入参数的错误，并提供了修正后的代码。

模型代码分析

class LeNet_1D_V(nn.Module):
    def __init__(self):
        super(LeNet_1D_V, self).__init__()
        self.conv1 = nn.Sequential(
            nn.Conv1d(1, 5, kernel_size=3, stride=1, padding=1),
            nn.Mish()
        )
        self.conv2 = nn.Sequential(
            nn.Conv1d(5, 10, kernel_size=3, stride=1, padding=1),
            nn.Mish()
        )
        self.conv3 = nn.Sequential(
            nn.Conv1d(10, 15, kernel_size=3, stride=1, padding=1),
            nn.Mish()
        )
        self.conv4 = nn.Sequential(
            nn.Conv1d(15, 20, kernel_size=3, stride=1, padding=1),
            nn.Mish()
        )
        self.conv5 = nn.Sequential(
            nn.Conv1d(20, 25, kernel_size=3, stride=1, padding=1),
            nn.Mish()
        )
        self.conv6 = nn.Sequential(
            nn.Conv1d(25, 30, kernel_size=3, stride=1, padding=1),
            nn.Mish()
        )
        self.conv7 = nn.Sequential(
            nn.Conv1d(30, 35, kernel_size=3, stride=1, padding=1),
            nn.Mish()
        )
        self.conv8 = nn.Sequential(
            nn.Conv1d(35, 40, kernel_size=3, stride=1, padding=1),
            nn.Mish()
        )
        self.avgpool = nn.AvgPool1d(kernel_size=2, stride=1)
        self.global_avgpool = nn.AdaptiveAvgPool1d(1)
        self.fc = nn.Linear(40, 6)
        self.softmax = nn.Softmax(dim=1)
        self.loss_fn = nn.CrossEntropyLoss()

    def forward(self, x):
        x = self.conv1(x)
        x = self.avgpool(x)
        x = self.conv2(x)
        x = self.avgpool(x)
        x = self.conv3(x)
        x = self.avgpool(x)
        x = self.conv4(x)
        x = self.avgpool(x)
        x = self.conv5(x)
        x = self.avgpool(x)
        x = self.conv6(x)
        x = self.avgpool(x)
        x = self.conv7(x)
        x = self.avgpool(x)
        x = self.conv8(x)
        x = self.global_avgpool(x)
        x = x.view(x.size(0), -1)
        feature = x
        x = self.fc(x)
        output = self.softmax(x)
        if target is not None:
            loss = self.loss_fn(output, target)
            return feature, output, loss
        else:
            return feature, output

model = LeNet_1D_V()
print(model)

模型结构：

包含八个卷积层（Conv）3*1
七个平均池化层（Avg-pool）2*1
一个全局平均池化层（GAP）
一个全连接层（FC）

模型特点：

第一层卷积核大小为5
激活函数为Mish函数（M）
多分类损失函数为交叉熵函数

代码错误分析

上述模型代码有一个错误，即在forward方法中缺少了target作为输入参数。

原因：

在代码中，forward方法的定义为forward(self, x)，但是后面却用到了target参数，导致程序报错。

代码修正

为了修正这个错误，需要将forward方法的定义改为forward(self, x, target)，并在方法中使用target参数。

class LeNet_1D_V(nn.Module):
    def __init__(self):
        super(LeNet_1D_V, self).__init__()
        self.conv1 = nn.Sequential(
            nn.Conv1d(1, 5, kernel_size=3, stride=1, padding=1),
            nn.Mish()
        )
        self.conv2 = nn.Sequential(
            nn.Conv1d(5, 10, kernel_size=3, stride=1, padding=1),
            nn.Mish()
        )
        self.conv3 = nn.Sequential(
            nn.Conv1d(10, 15, kernel_size=3, stride=1, padding=1),
            nn.Mish()
        )
        self.conv4 = nn.Sequential(
            nn.Conv1d(15, 20, kernel_size=3, stride=1, padding=1),
            nn.Mish()
        )
        self.conv5 = nn.Sequential(
            nn.Conv1d(20, 25, kernel_size=3, stride=1, padding=1),
            nn.Mish()
        )
        self.conv6 = nn.Sequential(
            nn.Conv1d(25, 30, kernel_size=3, stride=1, padding=1),
            nn.Mish()
        )
        self.conv7 = nn.Sequential(
            nn.Conv1d(30, 35, kernel_size=3, stride=1, padding=1),
            nn.Mish()
        )
        self.conv8 = nn.Sequential(
            nn.Conv1d(35, 40, kernel_size=3, stride=1, padding=1),
            nn.Mish()
        )
        self.avgpool = nn.AvgPool1d(kernel_size=2, stride=1)
        self.global_avgpool = nn.AdaptiveAvgPool1d(1)
        self.fc = nn.Linear(40, 6)
        self.softmax = nn.Softmax(dim=1)
        self.loss_fn = nn.CrossEntropyLoss()

    def forward(self, x, target):
        x = self.conv1(x)
        x = self.avgpool(x)
        x = self.conv2(x)
        x = self.avgpool(x)
        x = self.conv3(x)
        x = self.avgpool(x)
        x = self.conv4(x)
        x = self.avgpool(x)
        x = self.conv5(x)
        x = self.avgpool(x)
        x = self.conv6(x)
        x = self.avgpool(x)
        x = self.conv7(x)
        x = self.avgpool(x)
        x = self.conv8(x)
        x = self.global_avgpool(x)
        x = x.view(x.size(0), -1)
        feature = x
        x = self.fc(x)
        output = self.softmax(x)
        if target is not None:
            loss = self.loss_fn(output, target)
            return feature, output, loss
        else:
            return feature, output

model = LeNet_1D_V()
print(model)

现在，target作为输入参数传递给了forward方法。

总结

本文分析了LeNet-1D-V模型代码，并指出其在forward方法中缺少target输入参数的错误，并提供了修正后的代码。修正后的代码能够正常运行，并能够接收target参数，从而计算loss。