多通道二维卷积手动版+pytorch版本

1. 百度链接手动版

通过网盘分享的文件：conv2dtest.xlsx
链接: https://pan.baidu.com/s/1q3McqwfcKO1iX-Ms0BfAGA?pwd=ttsu 提取码: ttsu

2. Pytorch 版本

• python

import torch
import torch.nn as nn
import torch.nn.functional as F

torch.set_printoptions(precision=3, sci_mode=False)

if __name__ == "__main__":
    run_code = 0
    batch_size = 1
    in_channels = 2
    out_channels = 3
    input_h = 5
    input_w = 5
    total = batch_size * in_channels * input_h * input_w
    # input_matrix = torch.randn(batch_size, in_channels, input_h, input_w)
    input_matrix = torch.arange(total).reshape((batch_size, in_channels, input_h, input_w)).to(torch.float)
    kernel_size = 3
    # total = batch_size * in_channels * input_h * input_w
    print(f"input_matrix=\n{input_matrix}")
    my_conv2d = nn.Conv2d(out_channels=out_channels, in_channels=in_channels, kernel_size=kernel_size, stride=1,
                          bias=False)
    weight_0 = torch.arange(9).reshape((batch_size, kernel_size, kernel_size)).to(torch.float)
    weight_1 = torch.arange(9).reshape((batch_size, kernel_size, kernel_size)).to(torch.float) + 1
    weight_2 = torch.arange(9).reshape((batch_size, kernel_size, kernel_size)).to(torch.float) + 2
    weight_3 = torch.arange(9).reshape((batch_size, kernel_size, kernel_size)).to(torch.float) + 3
    weight_4 = torch.arange(9).reshape((batch_size, kernel_size, kernel_size)).to(torch.float) + 4
    weight_5 = torch.arange(9).reshape((batch_size, kernel_size, kernel_size)).to(torch.float) + 5
    weight_01 = torch.cat((weight_0, weight_1), dim=0)
    weight_23 = torch.cat((weight_2, weight_3), dim=0)
    weight_45 = torch.cat((weight_4, weight_5), dim=0)
    print(f"weight_01=\n{weight_01}")
    print(weight_01.shape)
    weight_01 = torch.unsqueeze(weight_01, dim=0)
    weight_23 = torch.unsqueeze(weight_23, dim=0)
    weight_45 = torch.unsqueeze(weight_45, dim=0)
    weight_012345 = torch.cat((weight_01, weight_23, weight_45), dim=0)
    print(f"weight_0123.shape=\n{weight_012345.shape}")
    print(f"weight_012345=\n{weight_012345}")
    my_conv2d.weight = nn.Parameter(weight_012345)
    output_matrix = my_conv2d(input_matrix)
    print(f"output_matrix=\n{output_matrix}")

• 结果：

input_matrix=
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., 24.]],

         [[25., 26., 27., 28., 29.],
          [30., 31., 32., 33., 34.],
          [35., 36., 37., 38., 39.],
          [40., 41., 42., 43., 44.],
          [45., 46., 47., 48., 49.]]]])
weight_01=
tensor([[[0., 1., 2.],
         [3., 4., 5.],
         [6., 7., 8.]],

        [[1., 2., 3.],
         [4., 5., 6.],
         [7., 8., 9.]]])
torch.Size([2, 3, 3])
weight_0123.shape=
torch.Size([3, 2, 3, 3])
weight_012345=
tensor([[[[ 0.,  1.,  2.],
          [ 3.,  4.,  5.],
          [ 6.,  7.,  8.]],

         [[ 1.,  2.,  3.],
          [ 4.,  5.,  6.],
          [ 7.,  8.,  9.]]],


        [[[ 2.,  3.,  4.],
          [ 5.,  6.,  7.],
          [ 8.,  9., 10.]],

         [[ 3.,  4.,  5.],
          [ 6.,  7.,  8.],
          [ 9., 10., 11.]]],


        [[[ 4.,  5.,  6.],
          [ 7.,  8.,  9.],
          [10., 11., 12.]],

         [[ 5.,  6.,  7.],
          [ 8.,  9., 10.],
          [11., 12., 13.]]]])
output_matrix=
tensor([[[[1803., 1884., 1965.],
          [2208., 2289., 2370.],
          [2613., 2694., 2775.]],

         [[2469., 2586., 2703.],
          [3054., 3171., 3288.],
          [3639., 3756., 3873.]],

         [[3135., 3288., 3441.],
          [3900., 4053., 4206.],
          [4665., 4818., 4971.]]]], grad_fn=<ConvolutionBackward0>)

• excel 版本结果：与pytorch版本一致，此版本无bias偏置。