# In[1] 导入所需工具包 import torch import torch.nn as nn import torchvision from torchvision import datasets, transforms import time from torch.nn import functional as F from math import floor, ceil from torch.utils.data import DataLoader,TensorDataset # import torchvision.transforms as transforms device = torch.device('cuda' if torch.cuda.is_available() else 'cpu') print(device) # In[1] 设置超参数 num_epochs = 60 batch_size = 1000 learning_rate = 0.001 # In[2] 获取数据包括训练数据和测试数据 transform = torchvision.transforms.Compose([ torchvision.transforms.ToTensor(), torchvision.transforms.Normalize( (0.1307,), (0.3081,)) ]) train_set = torchvision.datasets.MNIST(root='MNIST', train=True, download=True) train_data = train_set.data.float().unsqueeze(1) / 255.0 train_labels = train_set.targets train_dataset = TensorDataset(train_data,train_labels) train_loader = DataLoader(train_dataset,batch_size=batch_size,shuffle=True) test_set = torchvision.datasets.MNIST(root='MNIST', train=False, download=True) test_data = test_set.data.float().unsqueeze(1) / 255.0 test_labels = test_set.targets test_dataset = TensorDataset(test_data,test_labels) test_loader = DataLoader(test_dataset,batch_size=batch_size,shuffle=True) # In[1] 定义卷积核 def conv3x3(in_channels, out_channels, stride=1): return nn.Conv2d(in_channels, out_channels, kernel_size=3, stride=stride, padding=1, bias=True) # In[1] 定义残差块 class ResidualBlock(nn.Module): def __init__(self, in_channels, out_channels, stride=1, downsample=None): super(ResidualBlock, self).__init__() self.conv1 = conv3x3(in_channels, out_channels, stride) self.bn1 = nn.BatchNorm2d(out_channels) self.relu = nn.ReLU(inplace=True) self.conv2 = conv3x3(out_channels, out_channels) self.bn2 = nn.BatchNorm2d(out_channels) self.downsample = downsample def forward(self, x): residual = x out = self.conv1(x) out = self.bn1(out) out = self.relu(out) out = self.conv2(out) out = self.bn2(out) # 下采样 if self.downsample: residual = self.downsample(x) out += residual out = self.relu(out) return out # In[1] 搭建残差神经网络 class ResNet(nn.Module): def __init__(self, block, layers, num_classes=10): super(ResNet, self).__init__() self.in_channels = 16 self.conv = conv3x3(1, 16) self.bn = nn.BatchNorm2d(16) self.relu = nn.ReLU(inplace=True) # 构建残差块,恒等映射 # in_channels == out_channels and stride = 1 所以这里我们构建残差块,没有下采样 self.layer1 = self.make_layer(block, 16, layers[0], stride=1) # 不构建残差块,进行了下采样 # layers中记录的是数字,表示对应位置的残差块数目 self.layer2 = self.make_layer(block, 32, layers[1], 2) # 不构建残差块,进行了下采样 self.layer3 = self.make_layer(block, 64, layers[2], 2) self.avg_pool = nn.AvgPool2d(8) self.fc1 = nn.Linear(3136, 128) self.normfc12 = nn.LayerNorm((128), eps=1e-5) self.fc2 = nn.Linear(128, num_classes) def make_layer(self, block, out_channels, blocks, stride=1): downsample = None if (stride != 1) or (self.in_channels != out_channels): downsample = nn.Sequential( conv3x3(self.in_channels, out_channels, stride=stride), nn.BatchNorm2d(out_channels)) layers = [] layers.append(block(self.in_channels, out_channels, stride, downsample)) # 当out_channels = 32时,in_channels也变成32了 self.in_channels = out_channels # blocks是残差块的数目 # 残差块之后的网络结构,是out_channels->out_channels的 # 可以说,make_layer做的是输出尺寸相同的所有网络结构 # 由于输出尺寸会改变,我们用make_layers去生成一大块对应尺寸完整网络结构 for i in range(1, blocks): layers.append(block(out_channels, out_channels)) return nn.Sequential(*layers) def forward(self, x): out = self.conv(x) out = self.bn(out) out = self.relu(out) # layer1是三块in_channels等于16的网络结构，包括三个恒等映射 out = self.layer1(out) # layer2包括了16->32下采样,然后是32的三个恒等映射 out = self.layer2(out) # layer3包括了32->64的下采样,然后是64的三个恒等映射 out = self.layer3(out) # out = self.avg_pool(out) # 全连接压缩 # out.size(0)可以看作是batch_size out = out.view(out.size(0), -1) out = self.fc1(out) out = self.normfc12(out) out = self.relu(out) out = self.fc2(out) return out # In[1] 定义模型和损失函数 # [2,2,2]表示的是不同in_channels下的恒等映射数目 model = ResNet(ResidualBlock, [2, 2, 2]).to(device) criterion = nn.CrossEntropyLoss() optimizer = torch.optim.Adam(model.parameters(), lr=learning_rate) # In[1] 设置一个通过优化器更新学习率的函数 def update_lr(optimizer, lr): for param_group in optimizer.param_groups: param_group['lr'] = lr # In[1] 定义测试函数 def test(model, test_loader): model.eval() with torch.no_grad(): correct = 0 total = 0 for images, labels in test_loader: images = images.to(device) labels = labels.to(device) outputs = model(images) _, predicted = torch.max(outputs.data, 1) total += labels.size(0) correct += (predicted == labels).sum().item() print('Accuracy of the model on the test images: {} %'.format(100 * correct / total)) # In[1] 训练模型更新学习率 total_step = len(train_loader) curr_lr = learning_rate for epoch in range(num_epochs): in_epoch = time.time() for i, (images, labels) in enumerate(train_loader): images = images.to(device) labels = labels.to(device) # Forward pass outputs = model(images) loss = criterion(outputs, labels) # Backward and optimize optimizer.zero_grad() loss.backward() optimizer.step() if (i + 1) % 100 == 0: print("Epoch [{}/{}], Step [{}/{}] Loss: {:.4f}" .format(epoch + 1, num_epochs, i + 1, total_step, loss.item())) test(model, test_loader) out_epoch = time.time() print(f"use {(out_epoch - in_epoch) // 60}min{(out_epoch - in_epoch) % 60}s") if (epoch + 1) % 20 == 0: curr_lr /= 3 update_lr(optimizer, curr_lr) # In[1] 测试模型并保存 model.eval() with torch.no_grad(): correct = 0 total = 0 for images, labels in test_loader: images = images.to(device) labels = labels.to(device) outputs = model(images) _, predicted = torch.max(outputs.data, 1) total += labels.size(0) correct += (predicted == labels).sum().item() print('Accuracy of the model on the test images: {} %'.format(100 * correct / total)) torch.save(model.state_dict(), '../resnet.ckpt')