对x.view(x.size(0), -1)的一些理解

一般地，在CNN等网络中，都是通过卷积过滤器对目标进行计算，然而这些计算都是建立在高维数据。
最后，项目需要对数据进行分类或者识别，就需要全连接层Linear，这时候就需要将高维数据平铺变为低位数据。
Lenet 例子

class Lenet(nn.Module):
    def __init__(self):                            
        super(Lenet, self).__init__()
        ‘’‘卷积、池化处理’‘’
        self.conv = nn.Sequential(
            nn.Conv2d(1, 6, 3, stride=1, padding=1),
            nn.MaxPool2d(2, 2),
            nn.Conv2d(6, 16, 5, stride=1, padding=0),
            nn.MaxPool2d(2, 2)
        )
		‘’‘全连接层’‘’
        self.fc = nn.Sequential(                 
            nn.Linear(400, 120),
            nn.Linear(120, 84),
            nn.Linear(84, 10)
        )


    def forward(self, x):
        out = self.conv(x) 				#卷积网络（高维数据）
        out = out.view(out.size(0), -1)					#高维数据 ‘压’成 低维数据
        out = self.fc(out)					#全连接层（低维数据）
        return out

lenet = Lenet()

例子

import torch as t
a = t.arange(24)
a = a.reshape(4, 3, 2)

x = a.view(a.size(0), -1)
y = a.view(a.size(1), -1)
‘’‘输出对比’‘’
a = tensor([[[ 0,  1],
         [ 2,  3],
         [ 4,  5]],

        [[ 6,  7],
         [ 8,  9],
         [10, 11]],

        [[12, 13],
         [14, 15],
         [16, 17]],

        [[18, 19],
         [20, 21],
         [22, 23]]])
a.size is torch.Size([4, 3, 2])

x =  tensor([[ 0,  1,  2,  3,  4,  5],
        [ 6,  7,  8,  9, 10, 11],
        [12, 13, 14, 15, 16, 17],
        [18, 19, 20, 21, 22, 23]])
x.size is torch.Size([4, 6])

y =  tensor([[ 0,  1,  2,  3,  4,  5,  6,  7],
        [ 8,  9, 10, 11, 12, 13, 14, 15],
        [16, 17, 18, 19, 20, 21, 22, 23]])
y.size is torch.Size([3, 8])

ps：

x = a.view(a.size(0), -1)
等价于
x = x.view(x.size()[0], -1)