|
@@ -0,0 +1,58 @@
|
|
|
|
|
+from torch.nn import Module
|
|
|
|
|
+from torch.nn import Conv2d, Linear
|
|
|
|
|
+from torch.nn.functional import relu, max_pool2d
|
|
|
|
|
+import torch
|
|
|
|
|
+import torch.nn as nn
|
|
|
|
|
+class Lenet5(Module):
|
|
|
|
|
+ def __init__(self):
|
|
|
|
|
+ super(Lenet5, self).__init__()
|
|
|
|
|
+ # 第一层卷积层，输入通道为1，输出通道为6，卷积核大小为5
|
|
|
|
|
+ self.conv1 = nn.Conv2d(1, 6, kernel_size=5)
|
|
|
|
|
+ # 最大池化层，池化核大小为2，步长为2
|
|
|
|
|
+ self.pool = nn.MaxPool2d(kernel_size=2, stride=2)
|
|
|
|
|
+ # 第二层卷积层，输入通道为6，输出通道为16，卷积核大小为5
|
|
|
|
|
+ self.conv2 = nn.Conv2d(6, 16, kernel_size=5)
|
|
|
|
|
+ # 全连接层，输入节点数为16*4*4，输出节点数为120
|
|
|
|
|
+ self.fc1 = nn.Linear(16 * 4 * 4, 120)
|
|
|
|
|
+ # 全连接层，输入节点数为120，输出节点数为84
|
|
|
|
|
+ self.fc2 = nn.Linear(120, 84)
|
|
|
|
|
+ # 输出层，输入节点数为84，输出节点数为10
|
|
|
|
|
+ self.fc3 = nn.Linear(84, 10)
|
|
|
|
|
+
|
|
|
|
|
+ def forward(self, x):
|
|
|
|
|
+ # 第一层卷积，通过relu激活函数
|
|
|
|
|
+ x = self.pool(relu(self.conv1(x)))
|
|
|
|
|
+ # 第二层卷积，通过relu激活函数
|
|
|
|
|
+ x = self.pool(relu(self.conv2(x)))
|
|
|
|
|
+ # 展开张量，将其变成一维向量
|
|
|
|
|
+ x = x.view(-1, 16 * 4 * 4)
|
|
|
|
|
+ # 全连接层，通过relu激活函数
|
|
|
|
|
+ x = relu(self.fc1(x))
|
|
|
|
|
+ # 全连接层，通过relu激活函数
|
|
|
|
|
+ x = relu(self.fc2(x))
|
|
|
|
|
+ # 输出层，不使用激活函数
|
|
|
|
|
+ x = self.fc3(x)
|
|
|
|
|
+ return x
|
|
|
|
|
+
|
|
|
|
|
+
|
|
|
|
|
+
|
|
|
|
|
+# # 加载训练数据和测试数据
|
|
|
|
|
+# transform = transforms.Compose([
|
|
|
|
|
+# transforms.ToTensor(), # 转换为Tensor对象
|
|
|
|
|
+# transforms.Normalize((0.5,), (0.5,)) # 归一化处理
|
|
|
|
|
+# ])
|
|
|
|
|
+# trainset = torchvision.datasets.MNIST(root='./data', train=True, download=True, transform=transform)
|
|
|
|
|
+# trainloader = torch.utils.data.DataLoader(trainset, batch_size=64, shuffle=True)
|
|
|
|
|
+# testset = torchvision.datasets.MNIST(root='./data', train=False, download=True, transform=transform)
|
|
|
|
|
+# testloader = torch.utils.data.DataLoader(testset, batch_size=64, shuffle=False)
|
|
|
|
|
+#
|
|
|
|
|
+# # 实例化LeNet-5模型和损失函数、优化器
|
|
|
|
|
+# net = LeNet5()
|
|
|
|
|
+# criterion = nn.CrossEntropyLoss() # 交叉熵损失函数
|
|
|
|
|
+# optimizer = optim.SGD(net.parameters(), lr=0.001, momentum=0.9) # 随机梯度下降优化器
|
|
|
|
|
+#
|
|
|
|
|
+# # 训练网络
|
|
|
|
|
+# for epoch in range(10):
|
|
|
|
|
+# running_loss = 0.0
|
|
|
|
|
+# for i, data in enumerate(trainloader, 0):
|
|
|
|
|
+# inputs, labels = data
|
