PyTorch浅尝辄止(1)

与数据共舞

PyTorch跟数据相关的类有三个：torch.utils.data.DataLoader，torch.utils.data.Dataset，torch.utils.data.Sampler。Dataset是一个抽象类，你需要继承它并实现__len__和__getitem__方法。DataLoader是一个迭代器，它能够把Dataset包装起来，提供一些有用的方法。Sampler是一个抽象类，它能够提供索引的迭代器，用于DataLoader从Dataset中获取样本。

PyTorch跟领域相关的库包括torchaudio、torchvision和torchtext。torchaudio包含与音频处理相关的数据集。torchvision包含了一些常用的图像、视频数据集，比如MNIST，CIFAR10，COCO等。torchtext包含了一些常用的文本数据集，比如AG_NEWS，SogouNews，DBpedia等。

import torch
from torch import nn
from torch.utils.data import DataLoader
from torchvision import datasets
from torchvision.transforms import ToTensor

training_data = datasets.FashionMNIST(
    root="data",
    train=True,
    download=True,
    transform=ToTensor(),
)

test_data = datasets.FashionMNIST(
    root="data",
    train=False,
    download=True,
    transform=ToTensor(),
)

Downloading http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/train-images-idx3-ubyte.gz
Downloading http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/train-images-idx3-ubyte.gz to data\FashionMNIST\raw\train-images-idx3-ubyte.gz



  0%|          | 0/26421880 [00:00<?, ?it/s]


Extracting data\FashionMNIST\raw\train-images-idx3-ubyte.gz to data\FashionMNIST\raw

Downloading http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/train-labels-idx1-ubyte.gz
Downloading http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/train-labels-idx1-ubyte.gz to data\FashionMNIST\raw\train-labels-idx1-ubyte.gz



  0%|          | 0/29515 [00:00<?, ?it/s]


Extracting data\FashionMNIST\raw\train-labels-idx1-ubyte.gz to data\FashionMNIST\raw

Downloading http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/t10k-images-idx3-ubyte.gz
Downloading http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/t10k-images-idx3-ubyte.gz to data\FashionMNIST\raw\t10k-images-idx3-ubyte.gz



  0%|          | 0/4422102 [00:00<?, ?it/s]


Extracting data\FashionMNIST\raw\t10k-images-idx3-ubyte.gz to data\FashionMNIST\raw

Downloading http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/t10k-labels-idx1-ubyte.gz
Downloading http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/t10k-labels-idx1-ubyte.gz to data\FashionMNIST\raw\t10k-labels-idx1-ubyte.gz



  0%|          | 0/5148 [00:00<?, ?it/s]


Extracting data\FashionMNIST\raw\t10k-labels-idx1-ubyte.gz to data\FashionMNIST\raw

DataLoader是一个迭代器，它能够把Dataset包装起来，提供一些有用的方法。所以在调用时，把Dataset作为参数传递给DataLoader。后者会返回一个迭代器，我们可以用它来遍历Dataset，并支持自动分批、采样、打乱数据和多进程数据加载等功能。

下面我们定义批次大小为64，然后创建一个DataLoader实例，这样每次迭代就会返回64个样本。

batch_size = 64

# 创建一个 DataLoader 来迭代训练数据
train_dataloader = DataLoader(training_data, batch_size=batch_size)
test_dataloader = DataLoader(test_data, batch_size=batch_size)

for X, y in test_dataloader:
    print("Shape of X [N, C, H, W]: ", X.shape)
    print("Shape of y: ", y.shape, y.dtype)
    break

print(f"Label categories: {training_data.classes}, {len(training_data.classes)}")

print("Done!")

Shape of X [N, C, H, W]:  torch.Size([64, 1, 28, 28])
Shape of y:  torch.Size([64]) torch.int64
Label categories: ['T-shirt/top', 'Trouser', 'Pullover', 'Dress', 'Coat', 'Sandal', 'Shirt', 'Sneaker', 'Bag', 'Ankle boot'], 10
Done!

与模型同在

在PyTorch中定义神经网络，我们定义一个类，继承自nn.Module。在类的构造函数中，我们定义网络的结构，即各个层的类型和参数。在类的forward函数中，我们定义网络的前向传播逻辑。为了加速计算，当有硬件支持的时候，我们可以把网络的计算放在GPU上。

从上面对数据的形状的输出中，可以看到我们的数据批次为64，神经网络的输入为28*28，所以我们需要把数据展平，即把每个样本的形状从(1, 28, 28)变成(1, 784)。然后我们定义一个神经网络，它包含两个全连接层，第一个全连接层的输入为784，输出为512，第二个全连接层的输入为512，输出为10。

这个10是因为我们的数据集有10个类别，所以我们的输出也是10个类别的概率分布。

device =(
    "cuda"
    if torch.cuda.is_available()
    else "cpu"
)

print("Using {} device".format(device))

Using cpu device

# 定义模型
class NeuralNetwork(nn.Module):
    def __init__(self):
        super(NeuralNetwork, self).__init__()
        self.flatten = nn.Flatten()
        self.linear_relu_stack = nn.Sequential(
            nn.Linear(28*28, 512),
            nn.ReLU(),
            nn.Linear(512, 512),
            nn.ReLU(),
            nn.Linear(512, 10),
        )

    def forward(self, x):
        x = self.flatten(x)
        logits = self.linear_relu_stack(x)
        return logits

model = NeuralNetwork().to(device)
print(model)

NeuralNetwork(
  (flatten): Flatten(start_dim=1, end_dim=-1)
  (linear_relu_stack): Sequential(
    (0): Linear(in_features=784, out_features=512, bias=True)
    (1): ReLU()
    (2): Linear(in_features=512, out_features=512, bias=True)
    (3): ReLU()
    (4): Linear(in_features=512, out_features=10, bias=True)
  )
)

模型训练

在训练模型时，我们需要定义损失函数和优化器。损失函数用于衡量模型的预测结果和真实值之间的差异，优化器则用于更新模型的参数，使得模型的预测结果更接近真实值。

loss_fn = nn.CrossEntropyLoss()
optimizer = torch.optim.SGD(model.parameters(), lr=1e-3)

def train(dataloader, model, loss_fn, optimizer):
    size = len(dataloader.dataset)
    for batch, (X, y) in enumerate(dataloader):
        X, y = X.to(device), y.to(device)

        # Compute prediction error
        pred = model(X)
        loss = loss_fn(pred, y)

        # Backpropagation
        optimizer.zero_grad()
        loss.backward()
        optimizer.step()

        if batch % 100 == 0:
            loss, current = loss.item(), batch * len(X)
            print(f"loss: {loss:>7f}  [{current:>5d}/{size:>5d}]")

还应该在测试集上评估模型的性能，以确保模型在训练过程中没有过拟合。

def test(dataloader, model, loss_fn):
    size = len(dataloader.dataset)
    num_batches = len(dataloader)
    model.eval()
    test_loss, correct = 0, 0
    with torch.no_grad():
        for X, y in dataloader:
            X, y = X.to(device), y.to(device)
            pred = model(X)
            test_loss += loss_fn(pred, y).item()
            correct += (pred.argmax(1) == y).type(torch.float).sum().item()
    test_loss /= num_batches
    correct /= size
    print(
        f"Test Error: \n Accuracy: {(100*correct):>0.1f}%, Avg loss: {test_loss:>8f} \n"
    )

epochs = 5
for t in range(epochs):
    print(f"Epoch {t+1}\n-------------------------------")
    train(train_dataloader, model, loss_fn, optimizer)
    test(test_dataloader, model, loss_fn)
print("Done!")

Epoch 1
-------------------------------
loss: 2.302288  [    0/60000]
loss: 2.295353  [ 6400/60000]
loss: 2.275450  [12800/60000]
loss: 2.272285  [19200/60000]
loss: 2.248172  [25600/60000]
loss: 2.220821  [32000/60000]
loss: 2.220658  [38400/60000]
loss: 2.192120  [44800/60000]
loss: 2.190701  [51200/60000]
loss: 2.152144  [57600/60000]
Test Error: 
 Accuracy: 54.1%, Avg loss: 2.150920 

Epoch 2
-------------------------------
loss: 2.157440  [    0/60000]
loss: 2.149096  [ 6400/60000]
loss: 2.094668  [12800/60000]
loss: 2.113385  [19200/60000]
loss: 2.047285  [25600/60000]
loss: 1.991776  [32000/60000]
loss: 2.012696  [38400/60000]
loss: 1.941758  [44800/60000]
loss: 1.951613  [51200/60000]
loss: 1.859037  [57600/60000]
Test Error: 
 Accuracy: 58.6%, Avg loss: 1.871845 

Epoch 3
-------------------------------
loss: 1.904892  [    0/60000]
loss: 1.874474  [ 6400/60000]
loss: 1.766521  [12800/60000]
loss: 1.803286  [19200/60000]
loss: 1.680894  [25600/60000]
loss: 1.642499  [32000/60000]
loss: 1.653953  [38400/60000]
loss: 1.570671  [44800/60000]
loss: 1.598109  [51200/60000]
loss: 1.476388  [57600/60000]
Test Error: 
 Accuracy: 61.1%, Avg loss: 1.506833 

Epoch 4
-------------------------------
loss: 1.573699  [    0/60000]
loss: 1.539440  [ 6400/60000]
loss: 1.397835  [12800/60000]
loss: 1.462805  [19200/60000]
loss: 1.341058  [25600/60000]
loss: 1.344851  [32000/60000]
loss: 1.349764  [38400/60000]
loss: 1.286765  [44800/60000]
loss: 1.320334  [51200/60000]
loss: 1.216479  [57600/60000]
Test Error: 
 Accuracy: 63.6%, Avg loss: 1.243597 

Epoch 5
-------------------------------
loss: 1.317790  [    0/60000]
loss: 1.302650  [ 6400/60000]
loss: 1.139997  [12800/60000]
loss: 1.243143  [19200/60000]
loss: 1.119455  [25600/60000]
loss: 1.145241  [32000/60000]
loss: 1.161800  [38400/60000]
loss: 1.106545  [44800/60000]
loss: 1.146681  [51200/60000]
loss: 1.061042  [57600/60000]
Test Error: 
 Accuracy: 64.8%, Avg loss: 1.079930 

Done!

模型保存和读入

训练之后，我们可以保存模型的参数，以便后续使用。

torch.save(model.state_dict(), "model.pth")
print("Saved PyTorch Model State to model.pth")

Saved PyTorch Model State to model.pth

model_loaded = NeuralNetwork()
model_loaded.load_state_dict(torch.load("model.pth"))

<All keys matched successfully>

模型预测

直接使用训练好的模型可以进行预测。训练好的模型可以是在本地训练好的，也可以是从其他地方下载的。下面我们使用训练好的模型对测试集中的第一张图片进行预测。

classes = [
    "T-shirt/top",
    "Trouser",
    "Pullover",
    "Dress",
    "Coat",
    "Sandal",
    "Shirt",
    "Sneaker",
    "Bag",
    "Ankle boot",
]  # classes = train_dataset.classes,这里假装我们没有任何dataset
model_loaded.eval()
x, y = test_data[0][0], test_data[0][1]
with torch.no_grad():
    pred = model_loaded(x)
    predicted, actual = classes[pred[0].argmax(0)], classes[y]
    print(f'Predicted: "{predicted}", Actual: "{actual}"')

Predicted: "Ankle boot", Actual: "Ankle boot"

import matplotlib.pyplot as plt
plt.imshow(x.squeeze(), cmap='gray')
plt.title(f'Predicted: "{predicted}", Actual: "{actual}"')

Text(0.5, 1.0, 'Predicted: "Ankle boot", Actual: "Ankle boot"')

总结

1. PyTorch神经网络的概念相对直观；
2. 数据集合与数据集合装载器的概念；
3. 模型的训练、保存、读入、预测等。