torch.max_pool2d输入输出举例

Sure. torch.max_pool2d is a PyTorch function used for 2D Max pooling operation. The function takes as input a tensor and several arguments defining the pooling operation.

For example, consider the following input tensor:

x = tensor([[[1, 2, 3, 4],
            [5, 6, 7, 8],
            [9, 10, 11, 12],
            [13, 14, 15, 16]]])

The first argument 'x' is a 1x1x4x4 tensor. Now, let us consider the following max pooling operation:

y = torch.max_pool2d(x, kernel_size=2, stride=2)

Here, we have set the kernel_size and stride to 2. The output 'y' will be a 1x1x2x2 tensor obtained after 2x2 max pooling.

tensor([[[[ 6,  8],
          [14, 16]]]])

So, we can see that the max pooling operation selects the maximum value in a 2x2 window and returns it in the output tensor.