Containers:
(1) nn.Sequential:按顺序的包装多个网络层,常用于block构建
(2) nn.ModuleList:像python的list一样包装多个网络层(迭代性)
(3) nn.ModuleDict:像python的dict一样包装多个网络层(通过命名索引网络层)(索引性)
一、容器Sequential
nn.Sequential是nn.Module的容器,用于按顺序包装一组网络层
顺序性:个网络层级之间是严格按顺序构建
自带forward(): 自带的forward里,通过for循环依次执行前向传播运算
nn.Sequentia是nn.Module的容器,用于按顺序包装一组网络层,使这组网络层被看作使整体(或子模块)
Sequential继承自Module类,则有8个有序字典去管理它的属性
例如:对LeNet:
卷积、池化、卷积、池化包装成一个sequential,fc1、fc2、fc3也包装成一个sequential,将两个sequential组装起来就是LeNet
Sequential:
class LeNetSequential(nn.Module):
def __init__(self, classes):
#继承自Module,有8个有序字典
super(LeNetSequential,self).__init__()
self.classes = classes
self.features = nn.Sequential(
nn.Conv2d(3, 6, 5),
nn.ReLU(),
nn.MaxPool2d(kernel_size=2, stride=2)
nn.Conv2d(6, 16, 5),
nn.ReLU(),
nn.MaxPool2d(kernel_size=2, stride=2),)
self.classifier = nn.Sequential(
nn.Linear(16*5*5, 120),
nn.ReLU(),
nn.Linear(120, 84),
nn.ReLU(),
nn.Linear(84, self.classes),)
def forward(self, x):
x = self.features(x)
x = x.view(x.size()[0],-1)
x = self.classifier(x)
return x
#网络实例化(用的是__init__部分,构建子网络)
net = LeNetSequential(2)
#模拟图片
fake_img = torch.randn((4, 3, 32, 32), dtype=torch.float32)
#调用网络(用的是forward部分,拼接子网络)
out = net(fake_img)
class LeNetSequentialOrderDict(nn.Module):
def __init__(self, classes):
#继承自Module,有8个有序字典
super(LeNetSequentialOrderDict, self).__init__()
self.features = nn.Sequential(OrderDict({
'conv1': nn.Conv2d(3, 6, 5),
'relu1': nn.ReLU(),
'pool1': nn.Maxpool2d(kernel_size=2, stride=2),
'conv2': nn.Conv2d(6, 16, 5),
'relu2': nn.ReLU(),
'pool2': nn.Maxpool2d(kernel_size=2, stride=2),
}))
self.classifer = nn.Sequential(OrderDict({
'fc1': nn.Linear(16*5*5, 120),
'relu3': nn.ReLU(),
'fc2': nn.Linear(120, 84),
'relu4': nn.ReLU(),
'fc3': nn.Linear(84, classes),
}))
def forward(self, x):
x = self.features(x)
x = x.view(x.size()[0], -1)
x = self.classifer(x)
return x
#网络实例化(用的是__init__部分,构建子网络)
net = LeNetSequentialOrderDict(2)
#模拟图片
fake_img = torch.randn((4, 3, 32, 32), dtype=torch.float32)
#调用网络(用的是forward部分,拼接子网络)
out = net(fake_img)
二、容器ModuleList
nn.ModuleList是nn.Module的容器,用于包装一组网络层,以迭代的方式调用网络层:
append():在ModuleList后面添加网络层
extend():拼接两个ModuleList
insert():指定在ModuleList中位置插入网络层
class ModuleList(nn.Module):
def __init(self):
super(ModuleList, self).__init__()
self.linears = nn.ModuleList([nn.Linear(10,10)for i in range(20)])
def forward(self, x):
for i, linear in enumerate(self.linears):
x = linear(x)
return x
#网络实例化(用的是__init__部分,构建子网络)
net = ModuleList()
#输入的数据
fake_data = torch.ones((10, 10))
#调用网络
output = net(fake_data)
三、容器之ModuleDict
nn.容器之ModuleDict是nn.module的容器,用于包装一组网络层,以索引的方式调用网络层
主要方法:
clear():清空容器之ModuleDict
items():返回可迭代的键值对(key-value pairs)
keys():返回字典的键(key)
values():返回字典的值(value)
pop():返回一键值对,并从字典中删除
class ModuleDict(nn.Module):
def __init__(self):
super(ModuleDict, self).__init__()
self.choices = nn.ModuleDict({
'conv': nn.Conv2d(10, 10, 3),
'pool': nn.Maxpool2d(3)
})
self.activation = nn.ModuleDict({
'relu': nn.ReLU(),
'prelu': nn.PReLU()
})
def forward(self, x, choice, act):
x = self.choices[choice](x)
x = self.activation[act](x)
return x
#网络实例化(用的是__init__部分,构建子网络)
net = ModuleDict()
#输入数据
fake_img = torch.randn((4, 10, 32, 32))
#调用函数
output = net(fake_img, 'conv', 'relu')