pytorch AvgPool2d函数使用详解

import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.autograd import Variable
import numpy as np
 
 
 
input = Variable(torch.Tensor([[[1, 3, 3, 4, 5, 6, 7], [1, 2, 3, 4, 5, 6, 7]], [[1, 3, 3, 4, 5, 6, 7], [1, 2, 3, 4, 5, 6, 7]]]))
print("input shape",input.shape)
c = F.avg_pool1d(input, kernel_size=3, stride=2)
print(c)
print("c shape:",c.shape)
 
# m = nn.AvgPool2d(3, stride=2)
m = nn.AvgPool2d((2, 2), stride=(2, 2))
input = Variable(torch.randn(20, 18, 50, 32)) # bach是20,图片size是50*31,chanel是1８(通道是1８,也就是每张图有1８个fature map)
input = np.array([[[[1, 2, 3, 4], [1, 2, 3, 4], [1, 2, 3, 4], [1, 2, 3, 4]],
          [[1, 2, 3, 4], [1, 2, 3, 4], [1, 2, 3, 4], [1, 2, 3, 4]]],
         [[[1, 2, 3, 4], [1, 2, 3, 4], [1, 2, 3, 4], [1, 2, 3, 4]],
          [[1, 2, 3, 4], [1, 2, 3, 4], [1, 2, 3, 4], [1, 2, 3, 4]]]]) #size２*2*4*4
print("input shape:",input.shape)
input = Variable(torch.FloatTensor(input))
output = m(input)
print(output)
print("output shape:",output.shape)#(2,2,2,2)

input shape torch.Size([2, 2, 7])
tensor([[[ 2.3333, 4.0000, 6.0000],
     [ 2.0000, 4.0000, 6.0000]],
 
    [[ 2.3333, 4.0000, 6.0000],
     [ 2.0000, 4.0000, 6.0000]]])
c shape: torch.Size([2, 2, 3])
input shape: (2, 2, 4, 4)
tensor([[[[ 1.5000, 3.5000],
     [ 1.5000, 3.5000]],
 
     [[ 1.5000, 3.5000],
     [ 1.5000, 3.5000]]],
 
 
    [[[ 1.5000, 3.5000],
     [ 1.5000, 3.5000]],
 
     [[ 1.5000, 3.5000],
     [ 1.5000, 3.5000]]]])
output shape: torch.Size([2, 2, 2, 2])