DCGAN 的 Pytorch 权重初始化问题

Question

我正在尝试为 DCGAN 创建生成器并初始化自定义权重。在 Pytorch 教程中，代码如下：

# Generator Code

class Generator(nn.Module):
    def __init__(self, ngpu):
        super(Generator, self).__init__()
        self.ngpu = ngpu
        self.main = nn.Sequential(
            # input is Z, going into a convolution
            nn.ConvTranspose2d( nz, ngf * 8, 4, 1, 0, bias=False),
            nn.BatchNorm2d(ngf * 8),
            nn.ReLU(True),
            # state size. (ngf*8) x 4 x 4
            nn.ConvTranspose2d(ngf * 8, ngf * 4, 4, 2, 1, bias=False),
            nn.BatchNorm2d(ngf * 4),
            nn.ReLU(True),
            # state size. (ngf*4) x 8 x 8
            nn.ConvTranspose2d( ngf * 4, ngf * 2, 4, 2, 1, bias=False),
            nn.BatchNorm2d(ngf * 2),
            nn.ReLU(True),
            # state size. (ngf*2) x 16 x 16
            nn.ConvTranspose2d( ngf * 2, ngf, 4, 2, 1, bias=False),
            nn.BatchNorm2d(ngf),
            nn.ReLU(True),
            # state size. (ngf) x 32 x 32
            nn.ConvTranspose2d( ngf, nc, 4, 2, 1, bias=False),
            nn.Tanh()
            # state size. (nc) x 64 x 64
        )

    def forward(self, input):
        return self.main(input)

但是，我正在尝试为顺序操作创建块，并且我的生成器如下所示：

class ConvTr(nn.Module):
    def __init__(self, input_channels, output_channels, k_size, stride, pad, b = False):
        super().__init__()
        
        self.conv = nn.ConvTranspose2d(in_channels=input_channels,
                                       out_channels=output_channels,
                                       kernel_size=k_size,
                                       stride=stride,
                                       padding=pad,
                                       bias=b)
        
        self.batch_norm = nn.BatchNorm2d(input_channels)
        self.activation = nn.ReLU()
        
        
    def forward(self, x):
        x = self.conv(x)
        x = self.activation(x)
        x = self.batch_norm(x)
        return x



class Generator(nn.Module):
    
    def __init__(self, inputNoise=100):
        super().__init__()

        self.conv1 = ConvTr(input_channels = inputNoise, output_channels = 128, k_size = 4, stride = 2, pad = 0)
        self.conv2 = ConvTr(input_channels = 128, output_channels = 64, k_size = 3, stride = 2, pad = 1)
        self.conv3 = ConvTr(input_channels = 64, output_channels = 32, k_size = 4, stride = 2, pad = 1)
        self.conv4 = ConvTranspose2d(in_channels=32, out_channels=1, kernel_size=4, stride=2, padding=1)
        self.tanh = Tanh()
        

    def forward(self, x):
        
        x = self.conv1(x)
        x = self.conv2(x)
        x = self.conv3(x)
        x = self.conv4(x)
        output = self.tanh(x)
        return output

这里是权重初始化函数：

def weights_init(model):
    # get the class name
    classname = model.__class__.__name__
    # check if the classname contains the word "conv"
    if classname.find("Conv") != -1:
        # intialize the weights from normal distribution
        nn.init.normal_(model.weight.data, 0.0, 0.02)
    # otherwise, check if the name contains the word "BatcnNorm"
    elif classname.find("BatchNorm") != -1:
        # intialize the weights from normal distribution and set the
        # bias to 0
        nn.init.normal_(model.weight.data, 1.0, 0.02)
        nn.init.constant_(model.bias.data, 0)

现在当我尝试初始化权重时，出现以下错误。

raise AttributeError("'{}' object has no attribute '{}'".format(
AttributeError: 'ConvTr' object has no attribute 'weight'

Answer 1

三个问题：

1. 使用 model.apply 进行模块级操作（如初始化权重）
2. 使用 isinstance 找出它是哪一层
3. 不要使用

   .data

   ，它已经被弃用了很长时间，应该尽可能避免使用

要初始化权重，请执行以下操作

def init_weight(model):
    def _init_weight(module):
        if isinstance(module, nn.Conv2d):
            nn.init.normal_(module.weight, 0., 0.02)
        elif isinstance(module, nn.BatchNorm2d):
            nn.init.normal_(module.weight, 1.0, 0.02)
            nn.init.constant_(module.bias, 0)

    model.apply(_init_weight)

# m is the model you want to initialize
init_weight(m)