• pytorch-11.卷积神经网络（高级篇）

    Inception Moudel

     

    1. import torch
    2. from torchvision import transforms
    3. from torchvision import datasets
    4. from torch.utils.data import DataLoader
    5. import torch.nn.functional as F
    6. import torch.optim as optim
    7.  
    8. batch_size = 64
    9. transform = transforms.Compose([
    10.     transforms.ToTensor(),
    11.     transforms.Normalize((0.1307),(0.3081))
    12. ])
    13.  
    14. train_dataset = datasets.MNIST(root='../dataset/mnist/',train=True,download=True,transform=transform)
    15.  
    16. train_loader = DataLoader(train_dataset,shuffle = True,batch_size=batch_size)
    17.  
    18. test_dataset = datasets.MNIST(root='../dataset/mnist/',train=False,download=True,transform=transform)
    19.  
    20. test_loader = DataLoader(train_dataset,shuffle = False,batch_size=batch_size)
    21.  
    22. class InceptionA(torch.nn.Module):
    23.     def __init__(self,in_channels):
    24.         super(InceptionA, self).__init__()
    25.         self.branch1x1 = torch.nn.Conv2d(in_channels,16,kernel_size=1)
    26.  
    27.         self.branch5x5_1 = torch.nn.Conv2d(in_channels,16,kernel_size=1)
    28.         self.branch5x5_2 = torch.nn.Conv2d(16,24,kernel_size=5,padding=2)
    29.  
    30.         self.branch3x3_1 = torch.nn.Conv2d(in_channels, 16, kernel_size=1)
    31.         self.branch3x3_2 = torch.nn.Conv2d(16,24, kernel_size=3,padding=1)
    32.         self.branch3x3_3 = torch.nn.Conv2d(24, 24, kernel_size=3,padding=1)
    33.  
    34.         self.branch_pool = torch.nn.Conv2d(in_channels,24,kernel_size=1)
    35.  
    36.     def forward(self,x):
    37.         branch1x1 = self.branch1x1(x)
    38.  
    39.         branch5x5 = self.branch5x5_1(x)
    40.         branch5x5 = self.branch5x5_2(branch5x5)
    41.  
    42.         branch3x3 = self.branch3x3_1(x)
    43.         branch3x3 = self.branch3x3_2(branch3x3)
    44.         branch3x3 = self.branch3x3_3(branch3x3)
    45.  
    46.         branch_pool =F.avg_pool2d(x,kernel_size=3,stride=1,padding=1)
    47.         branch_pool = self.branch_pool(branch_pool)
    48.  
    49.         outputs = [branch1x1,branch5x5,branch3x3,branch_pool]
    50.         return torch.cat(outputs,dim=1)  #batch channel W H  dim=1即沿着channel的维度拼起来
    51.  
    52.  
    53. class Net(torch.nn.Module):
    54.     def __init__(self):
    55.         super(Net, self).__init__()
    56.         self.conv1 = torch.nn.Conv2d(1,10,kernel_size=5)
    57.         self.conv2 = torch.nn.Conv2d(88,20,kernel_size=5)  # input_channel , output_channel
    58.         self.incep1 = InceptionA(in_channels=10)
    59.         self.incep2 = InceptionA(in_channels=20)
    60.  
    61.         self.mp = torch.nn.MaxPool2d(2)
    62.         self.fc = torch.nn.Linear(1408,10)
    63.  
    64.  
    65.  
    66.     def forward(self,x):
    67.         batch_size = x.size(0)
    68.         x = F.relu(self.mp(self.conv1(x)))
    69.         x = self.incep1(x)
    70.         x = F.relu(self.mp(self.conv2(x)))
    71.         x = self.incep2(x)
    72.  
    73.         x = x.view(batch_size,-1)
    74.         x = self.fc(x)
    75.  
    76.         return x
    77.  
    78. model = Net()
    79. device = torch.device("cuda:0"if torch.cuda.is_available() else "cpu")
    80. model.to(device)
    81.  
    82. criterion = torch.nn.CrossEntropyLoss()
    83. optimizer = optim.SGD(model.parameters(),lr=0.01,momentum=0.5)
    84.  
    85. def train(epoch):
    86.     running_loss = 0.0
    87.     for batch_idx,data in enumerate(train_loader,0):
    88.         inputs, target = data
    89.         inputs, target = inputs.to(device),target.to(device) #送到GPU
    90.         optimizer.zero_grad()
    91.  
    92.         #forward + backward + update
    93.         outputs = model(inputs)
    94.         loss = criterion(outputs,target)
    95.         loss.backward()
    96.         optimizer.step()
    97.  
    98.         running_loss +=loss.item()
    99.         if batch_idx % 300 ==299:
    100.             print('[%d,%5d]loss:%.3f'%(epoch+1,batch_idx+1,running_loss/300))
    101.             running_loss = 0
    102.  
    103. def test():
    104.     correct = 0
    105.     total = 0
    106.     with torch.no_grad():
    107.         for data in test_loader:
    108.             images,labels = data
    109.             images, labels = images.to(device), labels.to(device)  # 送到GPU
    110.             outputs = model(images)
    111.             _,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
    112.             total +=labels.size(0)
    113.             correct +=(predicted==labels).sum().item()
    114.     print('Accuracy on test set:%d %%'%(100*correct/total) )
    115.  
    116.  
    117. if __name__ == '__main__':
    118.     for epoch in range(10):
    119.         train(epoch)
    120.         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

     

    1. import torch
    2. from torchvision import transforms
    3. from torchvision import datasets
    4. from torch.utils.data import DataLoader
    5. import torch.nn.functional as F
    6. import torch.optim as optim
    7.  
    8. batch_size = 64
    9. transform = transforms.Compose([
    10.     transforms.ToTensor(),
    11.     transforms.Normalize((0.1307),(0.3081))
    12. ])
    13.  
    14. train_dataset = datasets.MNIST(root='../dataset/mnist/',train=True,download=True,transform=transform)
    15.  
    16. train_loader = DataLoader(train_dataset,shuffle = True,batch_size=batch_size)
    17.  
    18. test_dataset = datasets.MNIST(root='../dataset/mnist/',train=False,download=True,transform=transform)
    19.  
    20. test_loader = DataLoader(train_dataset,shuffle = False,batch_size=batch_size)
    21.  
    22. class ResidualBlock(torch.nn.Module):
    23.     def __init__(self,channels):
    24.         super(ResidualBlock, self).__init__()
    25.         self.channels = channels
    26.         self.conv1 = torch.nn.Conv2d(channels,channels,kernel_size=3,padding=1)
    27.         self.conv2 = torch.nn.Conv2d(channels,channels,kernel_size=3,padding=1)
    28.  
    29.     def forward(self,x):
    30.         y = F.relu(self.conv1(x))
    31.         y = self.conv2(y)
    32.         return F.relu(x+y)
    33.  
    34.  
    35.  
    36. class Net(torch.nn.Module):
    37.     def __init__(self):
    38.         super(Net, self).__init__()
    39.         self.conv1 = torch.nn.Conv2d(1, 16, kernel_size=5)
    40.         self.conv2 = torch.nn.Conv2d(16, 32, kernel_size=5)  # input_channel , output_channel
    41.  
    42.         self.rblock1 = ResidualBlock(16)
    43.         self.rblock2 = ResidualBlock(32)
    44.  
    45.         self.mp = torch.nn.MaxPool2d(2)
    46.         self.fc = torch.nn.Linear(512, 10)
    47.  
    48.  
    49.  
    50.     def forward(self,x):
    51.         in_size = x.size(0)
    52.         x = self.mp(F.relu(self.conv1(x)))
    53.         x = self.rblock1(x)
    54.         x = self.mp(F.relu(self.conv2(x)))
    55.         x = self.rblock2(x)
    56.  
    57.         x = x.view(in_size,-1)
    58.         x = self.fc(x)
    59.  
    60.         return x
    61.  
    62. model = Net()
    63. device = torch.device("cuda:0"if torch.cuda.is_available() else "cpu")
    64. model.to(device)
    65.  
    66. criterion = torch.nn.CrossEntropyLoss()
    67. optimizer = optim.SGD(model.parameters(),lr=0.01,momentum=0.5)
    68.  
    69. def train(epoch):
    70.     running_loss = 0.0
    71.     for batch_idx,data in enumerate(train_loader,0):
    72.         inputs, target = data
    73.         inputs, target = inputs.to(device),target.to(device) #送到GPU
    74.         optimizer.zero_grad()
    75.  
    76.         #forward + backward + update
    77.         outputs = model(inputs)
    78.         loss = criterion(outputs,target)
    79.         loss.backward()
    80.         optimizer.step()
    81.  
    82.         running_loss +=loss.item()
    83.         if batch_idx % 300 ==299:
    84.             print('[%d,%5d]loss:%.3f'%(epoch+1,batch_idx+1,running_loss/300))
    85.             running_loss = 0
    86.  
    87. def test():
    88.     correct = 0
    89.     total = 0
    90.     with torch.no_grad():
    91.         for data in test_loader:
    92.             images,labels = data
    93.             images, labels = images.to(device), labels.to(device)  # 送到GPU
    94.             outputs = model(images)
    95.             _,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
    96.             total +=labels.size(0)
    97.             correct +=(predicted==labels).sum().item()
    98.     print('Accuracy on test set:%d %%'%(100*correct/total) )
    99.  
    100.  
    101. if __name__ == '__main__':
    102.     for epoch in range(10):
    103.         train(epoch)
    104.         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
     

  • 相关阅读:
    【光学】基于matlab迈克尔逊干涉仪动态仿真【含Matlab源码 2054期】
    .NET Emit 入门教程：第二部分：构建动态程序集（追加构建静态程序集教程）
    动态规划合集
    python 常见文件读取
    C++ extern关键字理解
    vue 常用指令
    常用设计模式——模板方法模式
    九、多项式朴素贝叶斯算法(Multinomial NB，Multinomial Naive Bayes)（有监督学习）
    Mycat分库分表
    使用API Monitor工具巧妙探测C++程序中监听某端口的模块
  • 原文地址：https://blog.csdn.net/m0_65188455/article/details/126215558