Pycharm中对Con2d函数的理解

nn.Conv2d(in_channels, out_channels, kernel_size, stride=1, padding=0, dilation=1, groups=1, bias=True，padding_mode='zero')

in_channels：输入信号通道个数

out_channels：输出信号通道个数

kernel_size：卷积核的大小

stride：步长

padding：输入的每一条边补充0的层数

dilation：卷积核默认的间距，默认为1

groups：从输入通道到输出通道的将数据分为几组，每组channel重用几次，out_channels/groups计算得到，其必须被输出通道和输入通道整除，默认为1

输出大小为：

import torch
import torch.nn as nn
a = nn.Conv2d(in_channels=3,out_channels=4,kernel_size=2,padding=0)
print(a)
print(list(a.parameters())[0].shape)
X = torch.rand((1,3,3,3))
print(X)
print(a(X))

得出的结果为

Conv2d(3, 4, kernel_size=(2, 2), stride=(1, 1))
torch.Size([4, 3, 2, 2])
tensor([[[[0.2443, 0.9249, 0.4239],
          [0.3013, 0.4009, 0.5954],
          [0.6660, 0.8274, 0.7630]],

         [[0.3257, 0.9352, 0.8601],
          [0.2355, 0.1280, 0.5163],
          [0.0043, 0.2029, 0.5315]],

         [[0.7454, 0.9867, 0.6015],
          [0.1719, 0.9849, 0.5443],
          [0.3487, 0.0569, 0.1718]]]])
tensor([[[[-0.5957, -0.3891],
          [-0.6881, -0.7487]],

         [[ 0.5373,  0.6412],
          [-0.0632,  0.3590]],

         [[ 0.6515,  0.5075],
          [ 0.5245,  0.6769]],

         [[ 0.2223,  0.4225],
          [ 0.2579,  0.1483]]]], grad_fn=)