PyTorch翻译官网教程1-QUICKSTART

官网链接

Quickstart — PyTorch Tutorials 2.0.1+cu117 documentation

快速开始

本节将介绍机器学习中常见任务的API。请参阅每个部分中的链接以深入了解。

数据处理

PyTorch有两个处理数据源，torch.utils.data.DataLoader 和 torch.utils.data.Dataset 。Dataset存储样本及其相应的标签，DataLoader在Dataset之上包装一个可迭代对象。

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

PyTorch提供了特定领域的库，如TorchText, TorchVision和TorchAudio，它们都包含数据集。在本教程中，我们将使用TorchVision数据集。

torchvision.datasets 模块包含了真实数据的Dataset对象。比如 CIFAR, COCO（完整列表参考Datasets — Torchvision 0.15 documentation）。在本教程中，我们使用FashionMNIST数据集。每个TorchVision数据集包含两个参数:transform和target_transform，分别用于修改样本和标签。

# Download training data from open datasets.
training_data = datasets.FashionMNIST(
    root="data",
    train=True,
    download=True,
    transform=ToTensor(),
)

# Download test data from open datasets.
test_data = datasets.FashionMNIST(
    root="data",
    train=False,
    download=True,
    transform=ToTensor(),
)

我们将Dataset作为DataLoader的入参，Dataset在数据集上包装了一个可迭代对象，并支持自动批处理、采样、洗牌和多进程数据加载。这里我们定义了一个批处理大小为64，即dataloader可迭代对象中的每个元素将返回一批64个特征和标签。

batch_size = 64

# Create data loaders.
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(f"Shape of X [N, C, H, W]: {X.shape}")
    print(f"Shape of y: {y.shape} {y.dtype}")
    break

输出：

Shape of X [N, C, H, W]: torch.Size([64, 1, 28, 28])
Shape of y: torch.Size([64]) torch.int64

---

创建模型

为了在PyTorch中定义一个神经网络，我们创建了一个继承了nn.Module的类，我们在__init__函数中定义网络层，并在forward函数中指定数据如何通过网络。为了加速神经网络的操作，我们将其转移到GPU或MPS(如果可用)。

注：

PyTorch MPS (Multi-Process Service)是 PyTorch 中的一种分布式训练方式。它是基于Apple的MPS(Metal Performance Shaders) 框架开发的

# Get cpu, gpu or mps device for training.
device = (
    "cuda"
    if torch.cuda.is_available()
    else "mps"
    if torch.backends.mps.is_available()
    else "cpu"
)
print(f"Using {device} device")

# Define model
class NeuralNetwork(nn.Module):
    def __init__(self):
        super().__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)

输出

Using mps device
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)
  )
)

注：因为我的设备是苹果，所以输出是mps,跟官网显示不同

---

优化模型参数

为了训练一个模型，我们需要一个损失函数和一个优化器。

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

在单个训练循环中，模型对训练数据集进行预测(批量提供给它)，并反向传播预测误差以调整模型的参数。

def train(dataloader, model, loss_fn, optimizer):
    size = len(dataloader.dataset)
    model.train()
    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
        loss.backward()
        optimizer.step()
        optimizer.zero_grad()

        if batch % 100 == 0:
            loss, current = loss.item(), (batch + 1) * len(X)
            print(f"loss: {loss:>7f}  [{current:>5d}/{size:>5d}]")

我们还根据测试数据集检查模型的性能，以确保它正在学习。

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")

训练过程在几个迭代(epochs)中进行。在每个epochs，模型学习参数以做出更好的预测。我们打印出模型在每个epochs的精度和损失;我们希望看到精度随着时间的推移而提高，损失随着时间的推移而减少。

输出

Epoch 1
-------------------------------
loss: 2.304695  [   64/60000]
loss: 2.293914  [ 6464/60000]
loss: 2.271139  [12864/60000]
loss: 2.267832  [19264/60000]
loss: 2.240983  [25664/60000]
loss: 2.217048  [32064/60000]
loss: 2.230957  [38464/60000]
loss: 2.190546  [44864/60000]
loss: 2.180454  [51264/60000]
loss: 2.167166  [57664/60000]
Test Error: 
 Accuracy: 43.7%, Avg loss: 2.148366 

Epoch 2
-------------------------------
loss: 2.155765  [   64/60000]
loss: 2.153187  [ 6464/60000]
loss: 2.084353  [12864/60000]
loss: 2.106838  [19264/60000]
loss: 2.051079  [25664/60000]
loss: 1.998855  [32064/60000]
loss: 2.030421  [38464/60000]
loss: 1.942099  [44864/60000]
loss: 1.941234  [51264/60000]
loss: 1.891874  [57664/60000]
Test Error: 
 Accuracy: 55.8%, Avg loss: 1.873747 

Epoch 3
-------------------------------
loss: 1.904033  [   64/60000]
loss: 1.885520  [ 6464/60000]
loss: 1.749947  [12864/60000]
loss: 1.801118  [19264/60000]
loss: 1.690538  [25664/60000]
loss: 1.652585  [32064/60000]
loss: 1.680197  [38464/60000]
loss: 1.571219  [44864/60000]
loss: 1.597052  [51264/60000]
loss: 1.505626  [57664/60000]
Test Error: 
 Accuracy: 61.0%, Avg loss: 1.510632 

Epoch 4
-------------------------------
loss: 1.579605  [   64/60000]
loss: 1.553953  [ 6464/60000]
loss: 1.388195  [12864/60000]
loss: 1.468328  [19264/60000]
loss: 1.347958  [25664/60000]
loss: 1.354385  [32064/60000]
loss: 1.368013  [38464/60000]
loss: 1.285745  [44864/60000]
loss: 1.321613  [51264/60000]
loss: 1.226315  [57664/60000]
Test Error: 
 Accuracy: 63.1%, Avg loss: 1.248957 

Epoch 5
-------------------------------
loss: 1.330482  [   64/60000]
loss: 1.320243  [ 6464/60000]
loss: 1.139326  [12864/60000]
loss: 1.250566  [19264/60000]
loss: 1.124903  [25664/60000]
loss: 1.160373  [32064/60000]
loss: 1.176003  [38464/60000]
loss: 1.108413  [44864/60000]
loss: 1.148409  [51264/60000]
loss: 1.063753  [57664/60000]
Test Error: 
 Accuracy: 64.3%, Avg loss: 1.086042 

Done!

---

保存模型

保存模型的一种常用方法是序列化内部状态字典(包含模型参数)。

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

输出

Saved PyTorch Model State to model.pth

---

加载模型

model = NeuralNetwork().to(device)
model.load_state_dict(torch.load("model.pth"))

这个模型现在可以用来做预测。

classes = [
    "T-shirt/top",
    "Trouser",
    "Pullover",
    "Dress",
    "Coat",
    "Sandal",
    "Shirt",
    "Sneaker",
    "Bag",
    "Ankle boot",
]

model.eval()
x, y = test_data[0][0], test_data[0][1]
with torch.no_grad():
    x = x.to(device)
    pred = model(x)
    predicted, actual = classes[pred[0].argmax(0)], classes[y]
    print(f'Predicted: "{predicted}", Actual: "{actual}"')

输出

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