Pytorch实现Resnet训练CIFAR10数据集（完整代码，可进一步优化）
Pytorch实现Resnet训练CIFAR10数据集（完整代码，包含resnet-50,resnet-101）
首先之前有写Pytorch的入门教程博客如果没有安装pytorch具体可转链接
废话不多说，直接上代码
这个代码使用CUDA 训练，如果不想使用GPU，可以将device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')注释掉，并且把所有后面用到的将tensor转到GPU的代码一起.to(device)删掉
CIFAR10数据集下载不动的这里有下载好的网盘链接（提取码5tk8），直接解压存入代码文件上一层的文件夹中
import torch
import torch.nn as nn
import torchvision
import torchvision.transforms as transforms
# 判断是否有GPU
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
num_epochs = 50 #50轮
batch_size = 50 #50步长
learning_rate = 0.01 #学习率0.01
# 图像预处理
transform = transforms.Compose([
transforms.Pad(4),
transforms.RandomHorizontalFlip(),
transforms.RandomCrop(32),
transforms.ToTensor()])
# CIFAR-10 数据集下载
train_dataset = torchvision.datasets.CIFAR10(root='../data/',
train=True,
transform=transform,
download=True)
test_dataset = torchvision.datasets.CIFAR10(root='../data/',
train=False,
transform=transforms.ToTensor())
# 数据载入
train_loader = torch.utils.data.DataLoader(dataset=train_dataset,
batch_size=batch_size,
shuffle=True)
test_loader = torch.utils.data.DataLoader(dataset=test_dataset,
batch_size=batch_size,
shuffle=False)
# 3x3 卷积定义
def conv3x3(in_channels, out_channels, stride=1):
return nn.Conv2d(in_channels, out_channels, kernel_size=3,
stride=stride, padding=1, bias=False)
# Resnet 的残差块
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)
if self.downsample:
residual = self.downsample(x)
out += residual
out = self.relu(out)
return out
# ResNet定义
class ResNet(nn.Module):
def __init__(self, block, layers, num_classes=10):
super(ResNet, self).__init__()
self.in_channels = 16
self.conv = conv3x3(3, 16)
self.bn = nn.BatchNorm2d(16)
self.relu = nn.ReLU(inplace=True)
self.layer1 = self.make_layer(block, 16, layers[0])
self.layer2 = self.make_layer(block, 32, layers[1], 2)
self.layer3 = self.make_layer(block, 64, layers[2], 2)
self.avg_pool = nn.AvgPool2d(8)
self.fc = nn.Linear(64, 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))
self.in_channels = out_channels
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)
out = self.layer1(out)
out = self.layer2(out)
out = self.layer3(out)
out = self.avg_pool(out)
out = out.view(out.size(0), -1)
out = self.fc(out)
return out
model = ResNet(ResidualBlock, [2, 2, 2]).to(device)
# 损失函数
criterion = nn.CrossEntropyLoss()
optimizer = torch.optim.Adam(model.parameters(), lr=learning_rate)
# 更新学习率
def update_lr(optimizer, lr):
for param_group in optimizer.param_groups:
param_group['lr'] = lr
# 训练数据集
total_step = len(train_loader)
curr_lr = learning_rate
for epoch in range(num_epochs):
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()))
# 延迟学习率
if (epoch+1) % 20 == 0:
curr_lr /= 3
update_lr(optimizer, curr_lr)
# 测试网络模型
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))
# S将模型保存
torch.save(model.state_dict(), 'resnet.ckpt')
目前跑了100轮准确率是88%
对于卷积部分，残差块部分都还有很大的优化空间
2021.12.1 更新
如下是更新的使用Resnet-50训练cifar10的代码，因为cifar10图片大小只有28*28使用resnet-50有点大材小用，如果还想把model改成resnet-101,只需要把代码中
model = ResNet(ResidualBlock, [3, 4, 6, 3]).to(device)
3，4，6，3 改为3，4，23，3
model = ResNet(ResidualBlock, [3, 4, 23, 3]).to(device)
下面是resnet-50的代码，安装pytorch直接运行即可跑
import math
import torch
import torch.nn as nn
import torchvision
import torchvision.transforms as transforms
# 判断是否有GPU
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
num_epochs = 5 # 50轮
batch_size = 50 # 50步长
learning_rate = 0.01 # 学习率0.01
#Cifar 数据集是32*32的图片，如若导入自己数据集记得修改图片宽高数值
img_height = 32
img_width = 32
# 图像预处理
transform = transforms.Compose([
transforms.Pad(4),
transforms.RandomHorizontalFlip(),
transforms.RandomCrop(32),
transforms.ToTensor()])
# CIFAR-10 数据集下载
train_dataset = torchvision.datasets.CIFAR10(root='data/',
train=True,
transform=transform,
download=True)
test_dataset = torchvision.datasets.CIFAR10(root='data/',
train=False,
transform=transforms.ToTensor())
# 数据载入
train_loader = torch.utils.data.DataLoader(dataset=train_dataset,
batch_size=batch_size,
shuffle=True)
test_loader = torch.utils.data.DataLoader(dataset=test_dataset,
batch_size=batch_size,
shuffle=False)
# 3x3 卷积定义
def conv3x3(in_channels, out_channels, kernel_size = 3,stride=1, padding=1):
return nn.Conv2d(in_channels, out_channels, kernel_size=kernel_size,
stride=stride, padding=padding, bias=False)
# Resnet_50 中的残差块
class ResidualBlock(nn.Module):
def __init__(self, in_channels, out_channels, stride=1, downsample=None):
super(ResidualBlock, self).__init__()
self.mid_channels = out_channels//4
self.conv1 = conv3x3(in_channels, self.mid_channels, kernel_size=1, stride=stride, padding=0)#Resnet50 中，从第二个残差块开始每个layer的第一个残差块都需要一次downsample
self.bn1 = nn.BatchNorm2d(self.mid_channels)
self.relu = nn.ReLU(inplace=True)
self.conv2 = conv3x3(self.mid_channels, self.mid_channels)
self.bn2 = nn.BatchNorm2d(self.mid_channels)
self.conv3 = conv3x3(self.mid_channels, out_channels,kernel_size=1,padding=0)
self.bn3 = nn.BatchNorm2d(out_channels)
self.downsample_0 = conv3x3(in_channels,out_channels,kernel_size=1,stride=1,padding=0)
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)
out = self.conv3(out)
out = self.bn3(out)
if self.downsample:
residual = self.downsample(x)
else:
residual = self.downsample_0(x)
out += residual
out = self.bn3(out)
out = self.relu(out)
return out
# ResNet定义
class ResNet(nn.Module):
def __init__(self, block, layers, num_classes=10):
super(ResNet, self).__init__()
self.conv = conv3x3(3, 64,kernel_size=7,stride=2)
self.bn = nn.BatchNorm2d(64)
self.max_pool = nn.MaxPool2d(3,2,padding=1)
self.layer1 = self.make_layer(block, 64, 256, layers[0])
self.layer2 = self.make_layer(block, 256, 512, layers[1], 2)
self.layer3 = self.make_layer(block, 512, 1024, layers[2], 2)
self.layer4 = self.make_layer(block, 1024, 2048, layers[3], 2)
self.avg_pool = nn.AvgPool2d(3,stride=1,padding=1)
self.fc = nn.Linear(math.ceil(img_height/32)*math.ceil(img_width/32)*2048, num_classes)
def make_layer(self, block, in_channels, out_channels, blocks, stride=1):
downsample = None
if (stride != 1):
downsample = nn.Sequential(
conv3x3(in_channels, out_channels, kernel_size=3,stride=stride),
nn.BatchNorm2d(out_channels))
layers = []
layers.append(block(in_channels, out_channels, stride, downsample))
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.max_pool(out)
out = self.layer1(out)
out = self.layer2(out)
out = self.layer3(out)
out = self.layer4(out)
out = self.avg_pool(out)
out = out.view( -1,math.ceil(img_height/32)*math.ceil(img_width/32)*2048)
return out
#Resnet-50 3-4-6-3 总计(3+4+6+3)*3=48 个conv层 加上开头的两个Conv 一共50层
model = ResNet(ResidualBlock, [3, 4, 6, 3]).to(device)
# 损失函数
criterion = nn.CrossEntropyLoss()
optimizer = torch.optim.Adam(model.parameters(), lr=learning_rate)
# 更新学习率
def update_lr(optimizer, lr):
for param_group in optimizer.param_groups:
param_group['lr'] = lr
# 训练数据集
total_step = len(train_loader)
curr_lr = learning_rate
for epoch in range(num_epochs):
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()))
# 延迟学习率
if (epoch + 1) % 20 == 0:
curr_lr /= 3
update_lr(optimizer, curr_lr)
# 测试网络模型
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))
# S将模型保存
torch.save(model.state_dict(), 'resnet.ckpt')