5.3 层

 

5.3.0 头文件

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

 

5.3.1 定义不带参数的层

# 定义不带参数的层
class CenteredLayer(nn.Module):
    def __init__(self):
        super().__init__()

    def forward(self, X):
        return X - X.mean()
layer = CenteredLayer()
Y = layer(torch.FloatTensor([1, 2, 3, 4, 5]))
print(Y)
# 输出:
# tensor([-2., -1.,  0.,  1.,  2.])

# 将层放到顺序块中
net = nn.Sequential(nn.Linear(8, 128), CenteredLayer())
Y = net(torch.rand(4, 8))
print(Y.mean())
# 输出:
# tensor(2.3283e-09, grad_fn=<MeanBackward0>)

 

5.3.2 定义带参数的层

# 定义不带参数的层
class MyLinear(nn.Module):
    # in_units:输入神经元个数
    # units:输出神经元个数
    def __init__(self, in_units, units):
        # 调用父类构造函数
        super().__init__()
        # 随机初始化权重w
        self.weight = nn.Parameter(torch.randn(in_units, units))
        # 随机初始化偏移量
        self.bias = nn.Parameter(torch.randn(units,))
    # 前向传播
    def forward(self, X):
        linear = torch.matmul(X, self.weight.data) + self.bias.data
        return F.relu(linear)
linear = MyLinear(5, 3)
print(linear.weight)
# 输出:
# tensor([[ 0.0612, -1.9050, -1.8448],
#         [ 0.5920, -0.0301, -0.9718],
#         [-1.1657,  0.3628,  0.4078],
#         [ 1.3419, -0.2535, -0.7558],
#         [-1.2800,  1.3049, -0.5029]], requires_grad=True)

 

 

本小节完整代码如下

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

# ------------------------------定义不带参数的层------------------------------------

# 定义不带参数的层
class CenteredLayer(nn.Module):
    def __init__(self):
        super().__init__()

    def forward(self, X):
        return X - X.mean()
layer = CenteredLayer()
Y = layer(torch.FloatTensor([1, 2, 3, 4, 5]))
print(Y)
# 输出:
# tensor([-2., -1.,  0.,  1.,  2.])

# 将层放到顺序块中
net = nn.Sequential(nn.Linear(8, 128), CenteredLayer())
Y = net(torch.rand(4, 8))
print(Y.mean())
# 输出:
# tensor(2.3283e-09, grad_fn=<MeanBackward0>)

# ------------------------------定义带参数的层------------------------------------

# 定义不带参数的层
class MyLinear(nn.Module):
    # in_units:输入神经元个数
    # units:输出神经元个数
    def __init__(self, in_units, units):
        # 调用父类构造函数
        super().__init__()
        # 随机初始化权重w
        self.weight = nn.Parameter(torch.randn(in_units, units))
        # 随机初始化偏移量
        self.bias = nn.Parameter(torch.randn(units,))
    # 前向传播
    def forward(self, X):
        linear = torch.matmul(X, self.weight.data) + self.bias.data
        return F.relu(linear)
linear = MyLinear(5, 3)
print(linear.weight)
# 输出:
# tensor([[ 0.0612, -1.9050, -1.8448],
#         [ 0.5920, -0.0301, -0.9718],
#         [-1.1657,  0.3628,  0.4078],
#         [ 1.3419, -0.2535, -0.7558],
#         [-1.2800,  1.3049, -0.5029]], requires_grad=True)