MINST数据集手写数字识别
学习课程:
吴恩达深度学习课程Deeplearning.ai
刘二大人PyTorch深度学习实践
1.导入模块
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
2.加载数据集
transform = transforms.Compose([
transforms.ToTensor(), # 将其转换为tensor类型
transforms.Normalize((0.1307,), (0.3081,)) # 将其压缩至0-1之间
])
# 加载数据准备
train_dataset = datasets.MNIST(root='../dataset/minst', train=True, transform=transform, download=True);
test_dataset = datasets.MNIST(root='../dataset/minst', train=False, transform=transform, download=True);
train_loader = DataLoader(dataset=train_dataset, batch_size=64, shuffle=True);
test_loader = DataLoader(dataset=train_dataset, batch_size=64, shuffle=False);
3.构建网络
# 构建网络
class Net(torch.nn.Module):
def __init__(self):
super(Net, self).__init__();
self.l1 = torch.nn.Linear(784, 512);
self.l2 = torch.nn.Linear(512, 256);
self.l3 = torch.nn.Linear(256, 128);
self.l4 = torch.nn.Linear(128, 64);
# 输出层
self.l5 = torch.nn.Linear(64, 10);
def forward(self, x):
x = x.view(-1, 784);
x = F.relu(self.l1(x));
x = F.relu(self.l2(x));
x = F.relu(self.l3(x));
x = F.relu(self.l4(x));
return self.l5(x);
4.损失函数、梯度下降
# 实例化网络
model = Net();
# 损失函数及其优化
# ‘多分类’交叉熵损失函数 CrossEntropyLoss() = softmax + log + NLLLoss() = log_softmax + NLLLoss()
loss1 = torch.nn.CrossEntropyLoss();
# 梯度下降
optimizer = optim.SGD(model.parameters(), lr=0.01, momentum=0.5);
5.训练模型
# 训练模型
def train(epoch):
loss_running = 0.0;
# enumerate返回值有两个,一个是序号,一个是数据(包含训练数据和标签)
for batch_index, data in enumerate(train_loader, 0):
batch_x, batch_y = data;
optimizer.zero_grad();
# 损失函数 前馈 反馈
output = model(batch_x); # output 就是y-hat
loss = loss1(output, batch_y);
loss.backward();
optimizer.step();
loss_running += loss.item();
if batch_index % 300 == 299:
print('[%d, %5d] loss: %.3f' % (epoch + 1, batch_index + 1, loss_running / 300))
loss_running = 0.0
6.测试模型
def test():
correct = 0;
total = 0;
with torch.no_grad():
for data in test_loader:
t_images, t_labels = data;
t_output = model(t_images);
_, predicted = torch.max(t_output, dim=1);
total += t_labels.size(0);
correct += (predicted == t_labels).sum().item();
print('Accuracy on test set: %d %%' % (100 * correct / total));
7.执行
if __name__ == '__main__':
for epoch in range(10):
train(epoch);
test();
8.保存模型
Path = "MINST_MY_Model.pt";
torch.save(model,Path);
saveModel = torch.load(Path);
saveModel.eval();