[pytorch] the most common function of view

problem

PyTorch Programming , Everywhere view, What does it mean ？

Method

view The core function of is to refactor Tensor Of size, The case is as follows ：

import torch

# view The function of is reconstruction tensor Of size

a = torch.randn(100,)
print(a.size()) #torch.Size([100])


b = a.view(5, -1)
print(b.size()) # torch.Size([5, 20])


c = a.view(2, 5, -1) # torch.Size([2, 5, 10])
print(c.size())

When we want to be lazy , When you don't want to calculate one of the dimensions , It can be used -1 Instead of , Be careful , Only lazy here 1 Dimensions .

The practical application

    def forward(self, x):
        x = self.feature(x)
        x = nn.AvgPool2d(2)(x)
        x = x.view(x.size(0), -1)
        y = self.classifier(x)
        return y

Conclusion

view It is widely used , Do you fully understand and master its usage ？