0

我为 PyTorch CNN 项目编写了自定义数据集和数据加载器。这是数据集的相关代码

class MyDataset(Dataset):

  def __init__(self): 
    pass

  def __len__(self):
    return COUNT

  def __getitem__(self, idx):
    x, y = X[idx], Y[idx]
    x = image_augment(x)  # custom func to resize image to 32x32
    return x, y

每次训练的形状x都是[4, 32, 32, 3].

这是我的网络代码,直接取自这个 PyTorch 示例

class Net(nn.Module):

    def __init__(self, nc):
        super(Net, self).__init__()
        self.conv1 = nn.Conv2d(3, 6, 5)
        self.pool = nn.MaxPool2d(2, 2)
        self.conv2 = nn.Conv2d(6, 16, 5)
        self.fc1 = nn.Linear(16 * 5 * 5, 120)
        self.fc2 = nn.Linear(120, 84)
        self.fc3 = nn.Linear(84, nc)

    def forward(self, x):
        x = self.pool(F.relu(self.conv1(x)))
        x = self.pool(F.relu(self.conv2(x)))
        x = x.view(-1, 16 * 5 * 5)
        x = F.relu(self.fc1(x))
        x = F.relu(self.fc2(x))
        x = self.fc3(x)
        return x

当我尝试用我的 DataLoader 中的数据训练这个网络时,我得到了错误语句Given groups=1, weight of size [6, 3, 5, 5], expected input[4, 32, 32, 3] to have 3 channels, but got 200 channels instead。在我看来,我的问题是来自我的 DataLoader 使用的数据的形状x.view(4, 3, 32, 32),但后来我收到一个错误,说我couldn't use Conv2D on a ByteTensor。我在这里有点迷路,非常感谢任何帮助。谢谢!

4

1 回答 1

0

我终于明白了。不得不x = x.view(x.shape[0], 3, self.img_height, self.img_width).type('torch.FloatTensor')。例如。这将使交换从[4, 32, 32, 3][4, 3, 32, 32]

于 2020-03-27T04:28:58.113 回答