将 'int' 转换为 pytorch 'Variable' 会出问题

Convert 'int' to pytorch 'Variable' makes problems

第一个使用 pytorch 的项目,我在尝试将 MNIST 标签 'int' 转换为火炬 'Variable' 时遇到了困难。调试器说它没有维度?!

# numpy mnist data
X_train, Y_train = read_data("training")
X_test , Y_test  = read_data("testing")

arr = np.zeros(5)
for i in range(5):
    # in your training loop:
    costs_ = 0
    for k in range(10000):
        optimizer.zero_grad()                            # zero the gradient buffers
        a = torch.from_numpy(np.expand_dims(X_train[k].flatten(), axis=0)).float()
        b = torch.from_numpy(np.array(Y_train[k], dtype=np.float)).float()
        input = Variable(a)
        output = net(input)
        target = Variable(b)                             # PROBLEM!!
        loss = criterion(output, target)
        loss.backward()
        optimizer.step()                                 # Does the update

        costs_ += loss.data.numpy()
    arr[i] = costs_ 
    print(i)

抛出的错误是:"RuntimeError: input and target have different number of elements: input[1 x 1] has 1 elements, while target[] has 0 elements at /b/wheel/pytorch-src/torch/lib/THNN/generic/MSECriterion.c:12"

错误告诉您到底发生了什么。您的 target 变量为空。

编辑(在下方评论后):

如果Y_train[k] = 5,则np.array(Y_train[k], dtype=np.float).shape = (),进而Variable(b)成为无量纲的张量。

为了解决这个问题,您需要将列表传递给 np.array() 而不是整数或浮点数。

像这样:

b = torch.from_numpy(np.array([Y_train[k]], dtype=np.float)).float()