4.4 nn.functional

1. nn.Module 和 nn.functional 不同之处

import torch as t
from torch.autograd import Variable as V
from torch import nn


input = V(t.randn(2,3))
model = nn.Linear(3,4)
output1 = model(input)
output2 = nn.functional.linear(input,model.weight,model.bias)

b1 = nn.functional.relu(input)
b2 = nn.ReLU()(input)
print(b1==b2)

2. 什么时候使用nn.Module，什么时候使用nn.functional

1. 如果模型有可学习的参数，最好使用nn.Module，否则既可以使用nn.functional，也可以使用nn.Module
2. 由于激活函数（ReLU、sigmoid、tanh）、池化（MaxPool）等层没有可学习参数，可以使用对应的functional函数代替
3. 而卷积、全连接层等具有可学习参数的网络，建议使用nn.Module

3. nn.Module和nn,functional搭配使用

import torch as t
from torch.autograd import Variable as V
from torch import nn
from torch.nn import functional as F

class Net(nn.Module):
    def __init__(self):
        super(Net,self).__init__()
        self.conv1 = nn.Conv2d(3,6,5)
        self.conv2 = nn.Conv2d(6,16,5)
        self.fc1 = nn.Linear(16*5*5,120)
        self.fc2 = nn.Linear(120,84)
        self.fc3 = nn.Linear(84,10)

    def forward(self, x):
        x = F.pool(F.relu(self.conv1(x)),2) # 池化窗口2*2
        x = F.pool(F.relu(self.conv2(x)),2)
        x = x.view(-1,16*5*5)
        x = F.relu(self.fc1)
        x = F.relu(self.fc2)
        x = self.fc3(x)
        return x