PyTorch 基础篇(4):前馈神经网络(Feedforward Neural Network)

PyTorch

PyTorch 基础篇(4):前馈神经网络(Feedforward Neural Network)

参考代码

yunjey的 pytorch tutorial系列

1
2
3
4
5
# 包
import torch
import torch.nn as nn
import torchvision
import torchvision.transforms as transforms
1
2
3
4
5
6
7
8
9
10
11
# 设备配置
# 有cuda就用cuda
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')

# 超参数设置
input_size = 784
hidden_size = 500
num_classes = 10
num_epochs = 5
batch_size = 100
learning_rate = 0.001
1
print(device)
cuda

MINIST 数据集加载

1
2
3
4
5
# 训练数据集
train_dataset = torchvision.datasets.MNIST(root='../../../data/minist',
train=True,
transform=transforms.ToTensor(),
download=True)
Downloading http://yann.lecun.com/exdb/mnist/train-images-idx3-ubyte.gz
Downloading http://yann.lecun.com/exdb/mnist/train-labels-idx1-ubyte.gz
Downloading http://yann.lecun.com/exdb/mnist/t10k-images-idx3-ubyte.gz
Downloading http://yann.lecun.com/exdb/mnist/t10k-labels-idx1-ubyte.gz
Processing...
Done!
1
2
3
4
# 测试数据集
test_dataset = torchvision.datasets.MNIST(root='../../../data/minist',
train=False,
transform=transforms.ToTensor())
1
2
3
4
5
6
7
8
9
10
# 数据加载器 Data Loader
# 训练数据加载器
train_loader = torch.utils.data.DataLoader(dataset=train_dataset,
batch_size=batch_size,
shuffle=True)

# 测试数据加载器
test_loader = torch.utils.data.DataLoader(dataset=test_dataset,
batch_size=batch_size,
shuffle=False)

自定义前馈神经网络

1
2
3
4
5
6
7
8
9
10
11
12
13
# 定义:有一个隐藏层的全连接的神经网络
class NeuralNet(nn.Module):
def __init__(self, input_size, hidden_size, num_classes):
super(NeuralNet, self).__init__()
self.fc1 = nn.Linear(input_size, hidden_size)
self.relu = nn.ReLU()
self.fc2 = nn.Linear(hidden_size, num_classes)

def forward(self, x):
out = self.fc1(x)
out = self.relu(out)
out = self.fc2(out)
return out
1
2
3
4
5
6
7
8
# 加载(实例化)一个网络模型
# to(device)可以用来将模型放在GPU上训练
model = NeuralNet(input_size, hidden_size, num_classes).to(device)

# 定义损失函数和优化器
# 再次,损失函数CrossEntropyLoss适合用于分类问题,因为它自带SoftMax功能
criterion = nn.CrossEntropyLoss()
optimizer = torch.optim.Adam(model.parameters(), lr=learning_rate)

训练模型

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
total_step = len(train_loader)

for epoch in range(num_epochs):
for i, (images, labels) in enumerate(train_loader):
# 将tensor移动到配置好的设备上(GPU)
images = images.reshape(-1, 28*28).to(device)
labels = labels.to(device)

# 前向传播
outputs = model(images)
loss = criterion(outputs, labels)

# 反向传播和优化
optimizer.zero_grad() # 还是要注意此处,每次迭代训练都需要清空梯度缓存
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()))
Epoch [1/5], Step [100/600], Loss: 0.3552
Epoch [1/5], Step [200/600], Loss: 0.2119
Epoch [1/5], Step [300/600], Loss: 0.3251
Epoch [1/5], Step [400/600], Loss: 0.2134
Epoch [1/5], Step [500/600], Loss: 0.1376
Epoch [1/5], Step [600/600], Loss: 0.1637
Epoch [2/5], Step [100/600], Loss: 0.1140
Epoch [2/5], Step [200/600], Loss: 0.2439
Epoch [2/5], Step [300/600], Loss: 0.1156
Epoch [2/5], Step [400/600], Loss: 0.0217
Epoch [2/5], Step [500/600], Loss: 0.0973
Epoch [2/5], Step [600/600], Loss: 0.1627
Epoch [3/5], Step [100/600], Loss: 0.1536
Epoch [3/5], Step [200/600], Loss: 0.0502
Epoch [3/5], Step [300/600], Loss: 0.0367
Epoch [3/5], Step [400/600], Loss: 0.1291
Epoch [3/5], Step [500/600], Loss: 0.0502
Epoch [3/5], Step [600/600], Loss: 0.0670
Epoch [4/5], Step [100/600], Loss: 0.0598
Epoch [4/5], Step [200/600], Loss: 0.0823
Epoch [4/5], Step [300/600], Loss: 0.0466
Epoch [4/5], Step [400/600], Loss: 0.0350
Epoch [4/5], Step [500/600], Loss: 0.0754
Epoch [4/5], Step [600/600], Loss: 0.0601
Epoch [5/5], Step [100/600], Loss: 0.0274
Epoch [5/5], Step [200/600], Loss: 0.0469
Epoch [5/5], Step [300/600], Loss: 0.1103
Epoch [5/5], Step [400/600], Loss: 0.0505
Epoch [5/5], Step [500/600], Loss: 0.0093
Epoch [5/5], Step [600/600], Loss: 0.0513

测试并保存模型

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
# 测试阶段为提高效率,可以不计算梯度
# 使用with torch.no_grad()函数

with torch.no_grad():
correct = 0
total = 0
for images, labels in test_loader:
images = images.reshape(-1, 28*28).to(device)
labels = labels.to(device)
outputs = model(images)
# 统计预测概率最大的下标
_, predicted = torch.max(outputs.data, 1)
total += labels.size(0)
correct += (predicted == labels).sum().item()

print('Accuracy of the network on the 10000 test images: {} %'.format(100 * correct / total))
Accuracy of the network on the 10000 test images: 97.47 %
1
2
# 保存模型
torch.save(model.state_dict(), 'model.ckpt')