PyTorch基础——torch.nn.Conv2d中自定义权重

torch.nn.Conv2d函数调用后会自动初始化weight和bias，本文主要涉及
如何自定义weight和bias为需要的数均分布类型:
   torch.nn.Conv2d.weight.data以及torch.nn.Conv2d.bias.data为torch.tensor类型，因此只要对这两个属性进行操作即可。

【sample】
以input_channels = 2, output_channels = 1 为例

In [1]: import torch
In [2]: import torch.nn as nn

In [3]: conv = nn.Conv2d(in_channels=2, out_channels=1, kernel_size=3)

# 此时weight以及bias已由nn.Conv2d初始化
conv.weight, conv.bias
Out[4]: 
(Parameter containing:
 tensor([[[[-0.0335,  0.0855, -0.0708],
           [-0.1672,  0.0902, -0.0077],
           [-0.0838, -0.1539, -0.0933]],
 
          [[-0.0496,  0.1807, -0.1477],
           [ 0.0397,  0.1963,  0.0932],
           [-0.2018, -0.0436,  0.1971]]]], requires_grad=True),
 Parameter containing:
 tensor([-0.1963], requires_grad=True))

# 手动设定
# conv.weight.data 以及 conv.bias.data属性为torch.tensor
# 因此只要获取conv.weight.data以及conv.bias.data属性，后续调用torch.tensor的不同方法即可进行修改
# 例如：全部修改为0
In [5]: conv.weight.data.zero_(), conv.bias.data.zero_()

In [6]: conv.weight, conv.bias
Out[6]: 
(Parameter containing:
 tensor([[[[0., 0., 0.],
           [0., 0., 0.],
           [0., 0., 0.]],
 
          [[0., 0., 0.],
           [0., 0., 0.],
           [0., 0., 0.]]]], requires_grad=True),
 Parameter containing:
 tensor([0.], requires_grad=True))