PyTorch一小時掌握之神經網絡分類篇

概述

對於 MNIST 手寫數據集的具體介紹, 我們在 TensorFlow 中已經詳細描述過, 在這裡就不多贅述. 有興趣的同學可以去看看之前的文章: https://www.jb51.net/article/222183.htm

在上一節的內容裡, 我們用 PyTorch 實現瞭回歸任務, 在這一節裡, 我們將使用 PyTorch 來解決分類任務.

導包

import torchvision
import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
import matplotlib.pyplot as plt

設置超參數

# 設置超參數
n_epochs = 3
batch_size_train = 64
batch_size_test = 1000
learning_rate = 0.01
momentum = 0.5
log_interval = 10
random_seed = 1
torch.manual_seed(random_seed)

讀取數據

# 數據讀取
train_loader = torch.utils.data.DataLoader(
    torchvision.datasets.MNIST('./data/', train=True, download=True,
                               transform=torchvision.transforms.Compose([
                                   torchvision.transforms.ToTensor(),
                                   torchvision.transforms.Normalize(
                                       (0.1307,), (0.3081,))
                               ])),
    batch_size=batch_size_train, shuffle=True)
    
test_loader = torch.utils.data.DataLoader(
    torchvision.datasets.MNIST('./data/', train=False, download=True,
                               transform=torchvision.transforms.Compose([
                                   torchvision.transforms.ToTensor(),
                                   torchvision.transforms.Normalize(
                                       (0.1307,), (0.3081,))
                               ])),
    batch_size=batch_size_test, shuffle=True)

examples = enumerate(test_loader)
batch_idx, (example_data, example_targets) = next(examples)

# 調試輸出
print(example_targets)
print(example_data.shape)

輸出結果:
tensor([7, 6, 7, 5, 6, 7, 8, 1, 1, 2, 4, 1, 0, 8, 4, 4, 4, 9, 8, 1, 3, 3, 8, 6,
2, 7, 5, 1, 6, 5, 6, 2, 9, 2, 8, 4, 9, 4, 8, 6, 7, 7, 9, 8, 4, 9, 5, 3,
1, 0, 9, 1, 7, 3, 7, 0, 9, 2, 5, 1, 8, 9, 3, 7, 8, 4, 1, 9, 0, 3, 1, 2,
3, 6, 2, 9, 9, 0, 3, 8, 3, 0, 8, 8, 5, 3, 8, 2, 8, 5, 5, 7, 1, 5, 5, 1,
0, 9, 7, 5, 2, 0, 7, 6, 1, 2, 2, 7, 5, 4, 7, 3, 0, 6, 7, 5, 1, 7, 6, 7,
2, 1, 9, 1, 9, 2, 7, 6, 8, 8, 8, 4, 6, 0, 0, 2, 3, 0, 1, 7, 8, 7, 4, 1,
3, 8, 3, 5, 5, 9, 6, 0, 5, 3, 3, 9, 4, 0, 1, 9, 9, 1, 5, 6, 2, 0, 4, 7,
3, 5, 8, 8, 2, 5, 9, 5, 0, 7, 8, 9, 3, 8, 5, 3, 2, 4, 4, 6, 3, 0, 8, 2,
7, 0, 5, 2, 0, 6, 2, 6, 3, 6, 6, 7, 9, 3, 4, 1, 6, 2, 8, 4, 7, 7, 2, 7,
4, 2, 4, 9, 7, 7, 5, 9, 1, 3, 0, 4, 4, 8, 9, 6, 6, 5, 3, 3, 2, 3, 9, 1,
1, 4, 4, 8, 1, 5, 1, 8, 8, 0, 7, 5, 8, 4, 0, 0, 0, 6, 3, 0, 9, 0, 6, 6,
9, 8, 1, 2, 3, 7, 6, 1, 5, 9, 3, 9, 3, 2, 5, 9, 9, 5, 4, 9, 3, 9, 6, 0,
3, 3, 8, 3, 1, 4, 1, 4, 7, 3, 1, 6, 8, 4, 7, 7, 3, 3, 6, 1, 3, 2, 3, 5,
9, 9, 9, 2, 9, 0, 2, 7, 0, 7, 5, 0, 2, 6, 7, 3, 7, 1, 4, 6, 4, 0, 0, 3,
2, 1, 9, 3, 5, 5, 1, 6, 4, 7, 4, 6, 4, 4, 9, 7, 4, 1, 5, 4, 8, 7, 5, 9,
2, 9, 4, 0, 8, 7, 3, 4, 2, 7, 9, 4, 4, 0, 1, 4, 1, 2, 5, 2, 8, 5, 3, 9,
1, 3, 5, 1, 9, 5, 3, 6, 8, 1, 7, 9, 9, 9, 9, 9, 2, 3, 5, 1, 4, 2, 3, 1,
1, 3, 8, 2, 8, 1, 9, 2, 9, 0, 7, 3, 5, 8, 3, 7, 8, 5, 6, 4, 1, 9, 7, 1,
7, 1, 1, 8, 6, 7, 5, 6, 7, 4, 9, 5, 8, 6, 5, 6, 8, 4, 1, 0, 9, 1, 4, 3,
5, 1, 8, 7, 5, 4, 6, 6, 0, 2, 4, 2, 9, 5, 9, 8, 1, 4, 8, 1, 1, 6, 7, 5,
9, 1, 1, 7, 8, 7, 5, 5, 2, 6, 5, 8, 1, 0, 7, 2, 2, 4, 3, 9, 7, 3, 5, 7,
6, 9, 5, 9, 6, 5, 7, 2, 3, 7, 2, 9, 7, 4, 8, 4, 9, 3, 8, 7, 5, 0, 0, 3,
4, 3, 3, 6, 0, 1, 7, 7, 4, 6, 3, 0, 8, 0, 9, 8, 2, 4, 2, 9, 4, 9, 9, 9,
7, 7, 6, 8, 2, 4, 9, 3, 0, 4, 4, 1, 5, 7, 7, 6, 9, 7, 0, 2, 4, 2, 1, 4,
7, 4, 5, 1, 4, 7, 3, 1, 7, 6, 9, 0, 0, 7, 3, 6, 3, 3, 6, 5, 8, 1, 7, 1,
6, 1, 2, 3, 1, 6, 8, 8, 7, 4, 3, 7, 7, 1, 8, 9, 2, 6, 6, 6, 2, 8, 8, 1,
6, 0, 3, 0, 5, 1, 3, 2, 4, 1, 5, 5, 7, 3, 5, 6, 2, 1, 8, 0, 2, 0, 8, 4,
4, 5, 0, 0, 1, 5, 0, 7, 4, 0, 9, 2, 5, 7, 4, 0, 3, 7, 0, 3, 5, 1, 0, 6,
4, 7, 6, 4, 7, 0, 0, 5, 8, 2, 0, 6, 2, 4, 2, 3, 2, 7, 7, 6, 9, 8, 5, 9,
7, 1, 3, 4, 3, 1, 8, 0, 3, 0, 7, 4, 9, 0, 8, 1, 5, 7, 3, 2, 2, 0, 7, 3,
1, 8, 8, 2, 2, 6, 2, 7, 6, 6, 9, 4, 9, 3, 7, 0, 4, 6, 1, 9, 7, 4, 4, 5,
8, 2, 3, 2, 4, 9, 1, 9, 6, 7, 1, 2, 1, 1, 2, 6, 9, 7, 1, 0, 1, 4, 2, 7,
7, 8, 3, 2, 8, 2, 7, 6, 1, 1, 9, 1, 0, 9, 1, 3, 9, 3, 7, 6, 5, 6, 2, 0,
0, 3, 9, 4, 7, 3, 2, 9, 0, 9, 5, 2, 2, 4, 1, 6, 3, 4, 0, 1, 6, 9, 1, 7,
0, 8, 0, 0, 9, 8, 5, 9, 4, 4, 7, 1, 9, 0, 0, 2, 4, 3, 5, 0, 4, 0, 1, 0,
5, 8, 1, 8, 3, 3, 2, 1, 2, 6, 8, 2, 5, 3, 7, 9, 3, 6, 2, 2, 6, 2, 7, 7,
6, 1, 8, 0, 3, 5, 7, 5, 0, 8, 6, 7, 2, 4, 1, 4, 3, 7, 7, 2, 9, 3, 5, 5,
9, 4, 8, 7, 6, 7, 4, 9, 2, 7, 7, 1, 0, 7, 2, 8, 0, 3, 5, 4, 5, 1, 5, 7,
6, 7, 3, 5, 3, 4, 5, 3, 4, 3, 2, 3, 1, 7, 4, 4, 8, 5, 5, 3, 2, 2, 9, 5,
8, 2, 0, 6, 0, 7, 9, 9, 6, 1, 6, 6, 2, 3, 7, 4, 7, 5, 2, 9, 4, 2, 9, 0,
8, 1, 7, 5, 5, 7, 0, 5, 2, 9, 5, 2, 3, 4, 6, 0, 0, 2, 9, 2, 0, 5, 4, 8,
9, 0, 9, 1, 3, 4, 1, 8, 0, 0, 4, 0, 8, 5, 9, 8])
torch.Size([1000, 1, 28, 28])

可視化展示

# 畫圖 (前6個)
fig = plt.figure()
for i in range(6):
    plt.subplot(2, 3, i + 1)
    plt.tight_layout()
    plt.imshow(example_data[i][0], cmap='gray', interpolation='none')
    plt.title("Ground Truth: {}".format(example_targets[i]))
    plt.xticks([])
    plt.yticks([])
plt.show()

輸出結果:

在這裡插入圖片描述

建立模型

# 創建model
class Net(nn.Module):
    def __init__(self):
        super(Net, self).__init__()
        self.conv1 = nn.Conv2d(1, 10, kernel_size=5)
        self.conv2 = nn.Conv2d(10, 20, kernel_size=5)
        self.conv2_drop = nn.Dropout2d()
        self.fc1 = nn.Linear(320, 50)
        self.fc2 = nn.Linear(50, 10)

    def forward(self, x):
        x = F.relu(F.max_pool2d(self.conv1(x), 2))
        x = F.relu(F.max_pool2d(self.conv2_drop(self.conv2(x)), 2))
        x = x.view(-1, 320)
        x = F.relu(self.fc1(x))
        x = F.dropout(x, training=self.training)
        x = self.fc2(x)
        return F.log_softmax(x)


network = Net()
optimizer = optim.SGD(network.parameters(), lr=learning_rate,
                      momentum=momentum)

訓練模型

# 訓練
train_losses = []
train_counter = []
test_losses = []
test_counter = [i * len(train_loader.dataset) for i in range(n_epochs + 1)]


def train(epoch):
    network.train()
    for batch_idx, (data, target) in enumerate(train_loader):
        optimizer.zero_grad()
        output = network(data)
        loss = F.nll_loss(output, target)
        loss.backward()
        optimizer.step()
        if batch_idx % log_interval == 0:
            print('Train Epoch: {} [{}/{} ({:.0f}%)]\tLoss: {:.6f}'.format(
                epoch, batch_idx * len(data), len(train_loader.dataset),
                       100. * batch_idx / len(train_loader), loss.item()))
            train_losses.append(loss.item())
            train_counter.append(
                (batch_idx * 64) + ((epoch - 1) * len(train_loader.dataset)))
            torch.save(network.state_dict(), './model.pth')
            torch.save(optimizer.state_dict(), './optimizer.pth')


def test():
    network.eval()
    test_loss = 0
    correct = 0
    with torch.no_grad():
        for data, target in test_loader:
            output = network(data)
            test_loss += F.nll_loss(output, target, size_average=False).item()
            pred = output.data.max(1, keepdim=True)[1]
            correct += pred.eq(target.data.view_as(pred)).sum()
    test_loss /= len(test_loader.dataset)
    test_losses.append(test_loss)
    print('\nTest set: Avg. loss: {:.4f}, Accuracy: {}/{} ({:.0f}%)\n'.format(
        test_loss, correct, len(test_loader.dataset),
        100. * correct / len(test_loader.dataset)))


for epoch in range(1, n_epochs + 1):
    train(epoch)
    test()

輸出結果:
Train Epoch: 1 [0/60000 (0%)] Loss: 2.297471
Train Epoch: 1 [6400/60000 (11%)] Loss: 1.934886
Train Epoch: 1 [12800/60000 (21%)] Loss: 1.242982
Train Epoch: 1 [19200/60000 (32%)] Loss: 0.979296
Train Epoch: 1 [25600/60000 (43%)] Loss: 1.277279
Train Epoch: 1 [32000/60000 (53%)] Loss: 0.721533
Train Epoch: 1 [38400/60000 (64%)] Loss: 0.759595
Train Epoch: 1 [44800/60000 (75%)] Loss: 0.469635
Train Epoch: 1 [51200/60000 (85%)] Loss: 0.422614
Train Epoch: 1 [57600/60000 (96%)] Loss: 0.417603

Test set: Avg. loss: 0.1988, Accuracy: 9431/10000 (94%)

Train Epoch: 2 [0/60000 (0%)] Loss: 0.277207
Train Epoch: 2 [6400/60000 (11%)] Loss: 0.328862
Train Epoch: 2 [12800/60000 (21%)] Loss: 0.396312
Train Epoch: 2 [19200/60000 (32%)] Loss: 0.301772
Train Epoch: 2 [25600/60000 (43%)] Loss: 0.253600
Train Epoch: 2 [32000/60000 (53%)] Loss: 0.217821
Train Epoch: 2 [38400/60000 (64%)] Loss: 0.395815
Train Epoch: 2 [44800/60000 (75%)] Loss: 0.265737
Train Epoch: 2 [51200/60000 (85%)] Loss: 0.323627
Train Epoch: 2 [57600/60000 (96%)] Loss: 0.236692

Test set: Avg. loss: 0.1233, Accuracy: 9622/10000 (96%)

Train Epoch: 3 [0/60000 (0%)] Loss: 0.500148
Train Epoch: 3 [6400/60000 (11%)] Loss: 0.338118
Train Epoch: 3 [12800/60000 (21%)] Loss: 0.452308
Train Epoch: 3 [19200/60000 (32%)] Loss: 0.374940
Train Epoch: 3 [25600/60000 (43%)] Loss: 0.323300
Train Epoch: 3 [32000/60000 (53%)] Loss: 0.203830
Train Epoch: 3 [38400/60000 (64%)] Loss: 0.379557
Train Epoch: 3 [44800/60000 (75%)] Loss: 0.334822
Train Epoch: 3 [51200/60000 (85%)] Loss: 0.361676
Train Epoch: 3 [57600/60000 (96%)] Loss: 0.218833

Test set: Avg. loss: 0.0911, Accuracy: 9723/10000 (97%)

完整代碼

import torchvision
import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
import matplotlib.pyplot as plt

# 設置超參數
n_epochs = 3
batch_size_train = 64
batch_size_test = 1000
learning_rate = 0.01
momentum = 0.5
log_interval = 100
random_seed = 1
torch.manual_seed(random_seed)

# 數據讀取
train_loader = torch.utils.data.DataLoader(
    torchvision.datasets.MNIST('./data/', train=True, download=True,
                               transform=torchvision.transforms.Compose([
                                   torchvision.transforms.ToTensor(),
                                   torchvision.transforms.Normalize(
                                       (0.1307,), (0.3081,))
                               ])),
    batch_size=batch_size_train, shuffle=True)

test_loader = torch.utils.data.DataLoader(
    torchvision.datasets.MNIST('./data/', train=False, download=True,
                               transform=torchvision.transforms.Compose([
                                   torchvision.transforms.ToTensor(),
                                   torchvision.transforms.Normalize(
                                       (0.1307,), (0.3081,))
                               ])),
    batch_size=batch_size_test, shuffle=True)

examples = enumerate(test_loader)
batch_idx, (example_data, example_targets) = next(examples)

# 調試輸出
print(example_targets)
print(example_data.shape)

# 畫圖 (前6個)
fig = plt.figure()
for i in range(6):
    plt.subplot(2, 3, i + 1)
    plt.tight_layout()
    plt.imshow(example_data[i][0], cmap='gray', interpolation='none')
    plt.title("Ground Truth: {}".format(example_targets[i]))
    plt.xticks([])
    plt.yticks([])
plt.show()


# 創建model
class Net(nn.Module):
    def __init__(self):
        super(Net, self).__init__()
        self.conv1 = nn.Conv2d(1, 10, kernel_size=5)
        self.conv2 = nn.Conv2d(10, 20, kernel_size=5)
        self.conv2_drop = nn.Dropout2d()
        self.fc1 = nn.Linear(320, 50)
        self.fc2 = nn.Linear(50, 10)

    def forward(self, x):
        x = F.relu(F.max_pool2d(self.conv1(x), 2))
        x = F.relu(F.max_pool2d(self.conv2_drop(self.conv2(x)), 2))
        x = x.view(-1, 320)
        x = F.relu(self.fc1(x))
        x = F.dropout(x, training=self.training)
        x = self.fc2(x)
        return F.log_softmax(x)


network = Net()
optimizer = optim.SGD(network.parameters(), lr=learning_rate,
                      momentum=momentum)

# 訓練
train_losses = []
train_counter = []
test_losses = []
test_counter = [i * len(train_loader.dataset) for i in range(n_epochs + 1)]


def train(epoch):
    network.train()
    for batch_idx, (data, target) in enumerate(train_loader):
        optimizer.zero_grad()
        output = network(data)
        loss = F.nll_loss(output, target)
        loss.backward()
        optimizer.step()
        if batch_idx % log_interval == 0:
            print('Train Epoch: {} [{}/{} ({:.0f}%)]\tLoss: {:.6f}'.format(
                epoch, batch_idx * len(data), len(train_loader.dataset),
                       100. * batch_idx / len(train_loader), loss.item()))
            train_losses.append(loss.item())
            train_counter.append(
                (batch_idx * 64) + ((epoch - 1) * len(train_loader.dataset)))
            torch.save(network.state_dict(), './model.pth')
            torch.save(optimizer.state_dict(), './optimizer.pth')


def test():
    network.eval()
    test_loss = 0
    correct = 0
    with torch.no_grad():
        for data, target in test_loader:
            output = network(data)
            test_loss += F.nll_loss(output, target, size_average=False).item()
            pred = output.data.max(1, keepdim=True)[1]
            correct += pred.eq(target.data.view_as(pred)).sum()
    test_loss /= len(test_loader.dataset)
    test_losses.append(test_loss)
    print('\nTest set: Avg. loss: {:.4f}, Accuracy: {}/{} ({:.0f}%)\n'.format(
        test_loss, correct, len(test_loader.dataset),
        100. * correct / len(test_loader.dataset)))


for epoch in range(1, n_epochs + 1):
    train(epoch)
    test()

到此這篇關於PyTorch一小時掌握之神經網絡分類篇的文章就介紹到這瞭,更多相關PyTorch神經網絡分類內容請搜索WalkonNet以前的文章或繼續瀏覽下面的相關文章希望大傢以後多多支持WalkonNet!

推薦閱讀: