Inception Moudel

import torch
from torchvision import transforms
from torchvision import datasets
from torch.utils.data import DataLoader
import torch.nn.functional as F
import torch.optim as optim

batch_size = 64
transform = transforms.Compose([
    transforms.ToTensor(),
    transforms.Normalize((0.1307),(0.3081))
])

train_dataset = datasets.MNIST(root='../dataset/mnist/',train=True,download=True,transform=transform)

train_loader = DataLoader(train_dataset,shuffle = True,batch_size=batch_size)

test_dataset = datasets.MNIST(root='../dataset/mnist/',train=False,download=True,transform=transform)

test_loader = DataLoader(train_dataset,shuffle = False,batch_size=batch_size)

class InceptionA(torch.nn.Module):
    def __init__(self,in_channels):
        super(InceptionA, self).__init__()
        self.branch1x1 = torch.nn.Conv2d(in_channels,16,kernel_size=1)

        self.branch5x5_1 = torch.nn.Conv2d(in_channels,16,kernel_size=1)
        self.branch5x5_2 = torch.nn.Conv2d(16,24,kernel_size=5,padding=2)

        self.branch3x3_1 = torch.nn.Conv2d(in_channels, 16, kernel_size=1)
        self.branch3x3_2 = torch.nn.Conv2d(16,24, kernel_size=3,padding=1)
        self.branch3x3_3 = torch.nn.Conv2d(24, 24, kernel_size=3,padding=1)

        self.branch_pool = torch.nn.Conv2d(in_channels,24,kernel_size=1)

    def forward(self,x):
        branch1x1 = self.branch1x1(x)

        branch5x5 = self.branch5x5_1(x)
        branch5x5 = self.branch5x5_2(branch5x5)

        branch3x3 = self.branch3x3_1(x)
        branch3x3 = self.branch3x3_2(branch3x3)
        branch3x3 = self.branch3x3_3(branch3x3)

        branch_pool =F.avg_pool2d(x,kernel_size=3,stride=1,padding=1)
        branch_pool = self.branch_pool(branch_pool)

        outputs = [branch1x1,branch5x5,branch3x3,branch_pool]
        return torch.cat(outputs,dim=1)  #batch channel W H  dim=1即沿着channel的维度拼起来


class Net(torch.nn.Module):
    def __init__(self):
        super(Net, self).__init__()
        self.conv1 = torch.nn.Conv2d(1,10,kernel_size=5)
        self.conv2 = torch.nn.Conv2d(88,20,kernel_size=5)  # input_channel , output_channel
        self.incep1 = InceptionA(in_channels=10)
        self.incep2 = InceptionA(in_channels=20)

        self.mp = torch.nn.MaxPool2d(2)
        self.fc = torch.nn.Linear(1408,10)



    def forward(self,x):
        batch_size = x.size(0)
        x = F.relu(self.mp(self.conv1(x)))
        x = self.incep1(x)
        x = F.relu(self.mp(self.conv2(x)))
        x = self.incep2(x)

        x = x.view(batch_size,-1)
        x = self.fc(x)

        return x

model = Net()
device = torch.device("cuda:0"if torch.cuda.is_available() else "cpu")
model.to(device)

criterion = torch.nn.CrossEntropyLoss()
optimizer = optim.SGD(model.parameters(),lr=0.01,momentum=0.5)

def train(epoch):
    running_loss = 0.0
    for batch_idx,data in enumerate(train_loader,0):
        inputs, target = data
        inputs, target = inputs.to(device),target.to(device) #送到GPU
        optimizer.zero_grad()

        #forward + backward + update
        outputs = model(inputs)
        loss = criterion(outputs,target)
        loss.backward()
        optimizer.step()

        running_loss +=loss.item()
        if batch_idx % 300 ==299:
            print('[%d,%5d]loss:%.3f'%(epoch+1,batch_idx+1,running_loss/300))
            running_loss = 0

def test():
    correct = 0
    total = 0
    with torch.no_grad():
        for data in test_loader:
            images,labels = data
            images, labels = images.to(device), labels.to(device)  # 送到GPU
            outputs = model(images)
            _,predicted = torch.max(outputs.data,dim=1)  #dim=1维度1，行是第0个维度，列是第1个维度
            total +=labels.size(0)
            correct +=(predicted==labels).sum().item()
    print('Accuracy on test set:%d %%'%(100*correct/total) )


if __name__ == '__main__':
    for epoch in range(10):
        train(epoch)
        test()

 结果：

[1,  300]loss:0.880
[1,  600]loss:0.198
[1,  900]loss:0.144
Accuracy on test set:96 %
[2,  300]loss:0.111
[2,  600]loss:0.099
[2,  900]loss:0.090
Accuracy on test set:97 %
[3,  300]loss:0.078
[3,  600]loss:0.071
[3,  900]loss:0.076
Accuracy on test set:98 %
[4,  300]loss:0.062
[4,  600]loss:0.065
[4,  900]loss:0.057
Accuracy on test set:98 %
[5,  300]loss:0.055
[5,  600]loss:0.057
[5,  900]loss:0.052
Accuracy on test set:98 %
[6,  300]loss:0.047
[6,  600]loss:0.049
[6,  900]loss:0.050
Accuracy on test set:98 %
[7,  300]loss:0.045
[7,  600]loss:0.046
[7,  900]loss:0.043
Accuracy on test set:98 %
[8,  300]loss:0.040
[8,  600]loss:0.037
[8,  900]loss:0.041
Accuracy on test set:98 %
[9,  300]loss:0.038
[9,  600]loss:0.038
[9,  900]loss:0.034
Accuracy on test set:99 %
[10,  300]loss:0.035
[10,  600]loss:0.034
[10,  900]loss:0.034
Accuracy on test set:99 %

Process finished with exit code 0
 

Residual Block

Residual Block解决了梯度消失的问题

import torch
from torchvision import transforms
from torchvision import datasets
from torch.utils.data import DataLoader
import torch.nn.functional as F
import torch.optim as optim

batch_size = 64
transform = transforms.Compose([
    transforms.ToTensor(),
    transforms.Normalize((0.1307),(0.3081))
])

train_dataset = datasets.MNIST(root='../dataset/mnist/',train=True,download=True,transform=transform)

train_loader = DataLoader(train_dataset,shuffle = True,batch_size=batch_size)

test_dataset = datasets.MNIST(root='../dataset/mnist/',train=False,download=True,transform=transform)

test_loader = DataLoader(train_dataset,shuffle = False,batch_size=batch_size)

class ResidualBlock(torch.nn.Module):
    def __init__(self,channels):
        super(ResidualBlock, self).__init__()
        self.channels = channels
        self.conv1 = torch.nn.Conv2d(channels,channels,kernel_size=3,padding=1)
        self.conv2 = torch.nn.Conv2d(channels,channels,kernel_size=3,padding=1)

    def forward(self,x):
        y = F.relu(self.conv1(x))
        y = self.conv2(y)
        return F.relu(x+y)



class Net(torch.nn.Module):
    def __init__(self):
        super(Net, self).__init__()
        self.conv1 = torch.nn.Conv2d(1, 16, kernel_size=5)
        self.conv2 = torch.nn.Conv2d(16, 32, kernel_size=5)  # input_channel , output_channel

        self.rblock1 = ResidualBlock(16)
        self.rblock2 = ResidualBlock(32)

        self.mp = torch.nn.MaxPool2d(2)
        self.fc = torch.nn.Linear(512, 10)



    def forward(self,x):
        in_size = x.size(0)
        x = self.mp(F.relu(self.conv1(x)))
        x = self.rblock1(x)
        x = self.mp(F.relu(self.conv2(x)))
        x = self.rblock2(x)

        x = x.view(in_size,-1)
        x = self.fc(x)

        return x

model = Net()
device = torch.device("cuda:0"if torch.cuda.is_available() else "cpu")
model.to(device)

criterion = torch.nn.CrossEntropyLoss()
optimizer = optim.SGD(model.parameters(),lr=0.01,momentum=0.5)

def train(epoch):
    running_loss = 0.0
    for batch_idx,data in enumerate(train_loader,0):
        inputs, target = data
        inputs, target = inputs.to(device),target.to(device) #送到GPU
        optimizer.zero_grad()

        #forward + backward + update
        outputs = model(inputs)
        loss = criterion(outputs,target)
        loss.backward()
        optimizer.step()

        running_loss +=loss.item()
        if batch_idx % 300 ==299:
            print('[%d,%5d]loss:%.3f'%(epoch+1,batch_idx+1,running_loss/300))
            running_loss = 0

def test():
    correct = 0
    total = 0
    with torch.no_grad():
        for data in test_loader:
            images,labels = data
            images, labels = images.to(device), labels.to(device)  # 送到GPU
            outputs = model(images)
            _,predicted = torch.max(outputs.data,dim=1)  #dim=1维度1，行是第0个维度，列是第1个维度
            total +=labels.size(0)
            correct +=(predicted==labels).sum().item()
    print('Accuracy on test set:%d %%'%(100*correct/total) )


if __name__ == '__main__':
    for epoch in range(10):
        train(epoch)
        test()

结果：

[1,  300]loss:0.499
[1,  600]loss:0.159
[1,  900]loss:0.123
Accuracy on test set:96 %
[2,  300]loss:0.089
[2,  600]loss:0.083
[2,  900]loss:0.080
Accuracy on test set:97 %
[3,  300]loss:0.064
[3,  600]loss:0.062
[3,  900]loss:0.062
Accuracy on test set:98 %
[4,  300]loss:0.052
[4,  600]loss:0.052
[4,  900]loss:0.050
Accuracy on test set:98 %
[5,  300]loss:0.042
[5,  600]loss:0.043
[5,  900]loss:0.044
Accuracy on test set:98 %
[6,  300]loss:0.035
[6,  600]loss:0.040
[6,  900]loss:0.037
Accuracy on test set:98 %
[7,  300]loss:0.033
[7,  600]loss:0.033
[7,  900]loss:0.033
Accuracy on test set:98 %
[8,  300]loss:0.029
[8,  600]loss:0.029
[8,  900]loss:0.031
Accuracy on test set:99 %
[9,  300]loss:0.024
[9,  600]loss:0.028
[9,  900]loss:0.026
Accuracy on test set:99 %
[10,  300]loss:0.025
[10,  600]loss:0.023
[10,  900]loss:0.025
Accuracy on test set:99 %

Process finished with exit code 0