Resnet实现：forward接受1个位置参数，但给出了2个

如何解决Resnet实现：forward接受1个位置参数，但给出了2个

我编写了这段代码，当我运行它时，出现以下错误：forward（）接受1个位置参数，但给出了2个。据我所知，我仅将一个参数传递给forward（）。

• ResNet是基本的残差块

class ResNet(nn.Module):
    def __init__(self,in_channels,mid_channels,mid2_channels,out_channels):
        super().__init__()
        self.conv1 = nn.Conv2d(in_channels,kernel_size = 3,stride = 1,padding = 1)
        self.conv1_bn = nn.BatchNorm2d(mid_channels)
        self.conv2 = nn.Conv2d(mid_channels,padding = 1)
        self.conv2_bn = nn.BatchNorm2d(mid2_channels)
        self.conv3 = nn.Conv2d(mid2_channels,out_channels,padding = 1)
        self.conv3_bn = nn.BatchNorm2d(out_channels)
        if (in_channels != out_channels):
            self.conv_shortcut = nn.Conv2d(in_channels,kernel_size = 1,padding = 0 )

    def forward(self,X):
         X_shortcut = X
        X = F.relu(self.conv1(X))
        X =  self.conv1_bn(X)
        X = F.relu(self.conv2(X))
        X =  self.conv2_bn(X)
        X = F.relu(self.conv2(X))
        X =  self.conv2_bn(X)
        if (in_channels == out_channels):
            X = self.conv3(X) + X_shortcut
        else:
            X = self.conv3(X) + self.conv_shortcut(X_shortcut)
        X = self.conv3_bn(F.relu(x))
        return X

• 这是使用给定图层生成模型的方法。

class TotalNet(nn.Module):
    def __init__(self,Layers):
        super().__init__()
        self.hidden = nn.ModuleList()
        self.hidden.append(nn.BatchNorm2d(1))
        for i in range(0,len(Layers)-1,3):
            in_channels,out_channels = Layers[i:(i+4)]
            self.hidden.append(ResNet(in_channels,out_channels))
            self.hidden.append(nn.Flatten())
            
    def forward(self,X):
        X = self.hidden(X)
        return X

• 以下是我调用该函数的方式：

test = TotalNet([9,2,9,9])
a = torch.rand((1,9),dtype = torch.float32)
test(a)

解决方法

我意识到我正在将X传递给nn.ModuleList。正确的方法是将X应用于nn.ModuleList的元素并更新X的值，这是不正确的。 换句话说，TotalNet的转发功能应为：

for operation in self.hidden:
            X = operation(X)
        return X