1

我有一个预定义 nz=10, ngf= 64 的 python 测试代码

def test_Generator_shapes():
    nz = 10
    netG = Generator(nz, ngf=64, nc=1)

    batch_size = 32
    noise = torch.randn(batch_size, nz, 1, 1)
    out = netG(noise, verbose=True)

    assert out.shape == torch.Size([batch_size, 1, 28, 28]), f"Bad shape of out: out.shape={out.shape}"
    print('Success')

test_Generator_shapes()

现在我需要重置隐藏层和其他参数才能输出大小为 28x28 的图像，即 -torch.Size([batch_size, 1, 28, 28])

请有人建议我应该在以下代码中进行哪些更改，以便能够生成 28x28 而不是 64x64 目前的图像

class Generator(nn.Module):
    def __init__(self, nz=10, ngf=28, nc=1, ndf=28):
        """GAN generator.
        
        Args:
          nz:  Number of elements in the latent code.
          ngf: Base size (number of channels) of the generator layers.
          nc:  Number of channels in the generated images.
        """
        ngf=28
        super(Generator, self).__init__()
        self.ngpu = 0
        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
        )
        # YOUR CODE HERE
        #raise NotImplementedError()

    def forward(self, z, verbose=False):
        """Generate images by transforming the given noise tensor.
        
        Args:
          z of shape (batch_size, nz, 1, 1): Tensor of noise samples. We use the last two singleton dimensions
                          so that we can feed z to the generator without reshaping.
          verbose (bool): Whether to print intermediate shapes (True) or not (False).
        
        Returns:
          out of shape (batch_size, nc, 28, 28): Generated images.
        """
        # YOUR CODE HERE
        x = self.main(z)
        print (x.size())
        return x
        #raise NotImplementedError()
4

0 回答 0