PyTorch快速开始
创建于 2024年7月27日修改于 2024年7月27日
本文带你速览一遍 PyTorch 机器学习任务的常用API。想要深入了解，可点击各节提供的链接。
Contents
处理数据
PyTorch 提供操作数据的两个原语操作：torch.utils.data.DataLoader 和 torch.utils.data.Dataset。
Dataset 存储样本及其对应的标签，DataLoader 为数据集 Dataset 包装了一个可迭代对象。
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，例如 TorchText、TorchVision 和 TorchAudio，这些库都自带数据集。本教程将使用 TorchVision 数据集。
torchvision.datasets 模块包含许多业界视觉数据的 Dataset 对象，例如 CIFAR、COCO（完整列表在此）。在本教程中，我们使用 FashionMNIST 数据集。每个 TorchVision 数据集都包括两个参数：transform 和 target_transform，分别用于修改样本（samples）和标签（labels）。
# 从开放数据集中下载训练数据
training_data = datasets.FashionMNIST(
root="data",
train=True,
download=True,
transform=ToTensor(),
)
# 从开放数据集中下载验证测试数据
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]
0%| | 32768/26421880 [00:00<02:24, 182945.57it/s]
0%| | 65536/26421880 [00:00<02:25, 181484.75it/s]
0%| | 98304/26421880 [00:00<02:25, 180977.88it/s]
0%| | 131072/26421880 [00:00<02:25, 180861.96it/s]
1%| | 196608/26421880 [00:00<02:05, 208200.89it/s]
1%| | 262144/26421880 [00:01<01:42, 254132.99it/s]
1%|1 | 294912/26421880 [00:01<01:52, 232313.95it/s]
1%|1 | 360448/26421880 [00:01<01:36, 270874.17it/s]
2%|1 | 425984/26421880 [00:01<01:27, 297669.11it/s]
2%|1 | 491520/26421880 [00:01<01:21, 317096.10it/s]
2%|2 | 589824/26421880 [00:02<01:07, 383592.80it/s]
2%|2 | 655360/26421880 [00:02<01:08, 377776.82it/s]
3%|2 | 753664/26421880 [00:02<01:00, 426529.09it/s]
3%|3 | 851968/26421880 [00:02<00:55, 461764.06it/s]
4%|3 | 983040/26421880 [00:02<00:47, 538213.56it/s]
4%|4 | 1114112/26421880 [00:02<00:42, 592530.18it/s]
5%|4 | 1245184/26421880 [00:03<00:39, 632474.48it/s]
5%|5 | 1376256/26421880 [00:03<00:35, 714894.16it/s]
6%|5 | 1474560/26421880 [00:03<00:34, 716158.63it/s]
6%|6 | 1638400/26421880 [00:03<00:29, 847754.23it/s]
7%|6 | 1736704/26421880 [00:03<00:30, 818063.83it/s]
7%|7 | 1966080/26421880 [00:03<00:25, 965142.80it/s]
8%|8 | 2162688/26421880 [00:04<00:23, 1011318.88it/s]
9%|9 | 2424832/26421880 [00:04<00:20, 1145733.67it/s]
10%|# | 2686976/26421880 [00:04<00:19, 1244417.66it/s]
11%|#1 | 2981888/26421880 [00:04<00:17, 1362155.80it/s]
13%|#2 | 3309568/26421880 [00:04<00:14, 1614458.43it/s]
13%|#3 | 3506176/26421880 [00:04<00:14, 1572814.00it/s]
15%|#4 | 3899392/26421880 [00:05<00:12, 1771000.86it/s]
16%|#6 | 4292608/26421880 [00:05<00:10, 2059984.16it/s]
17%|#7 | 4554752/26421880 [00:05<00:10, 2024741.07it/s]
19%|#8 | 5013504/26421880 [00:05<00:08, 2398785.61it/s]
20%|## | 5308416/26421880 [00:05<00:09, 2340504.89it/s]
22%|##2 | 5865472/26421880 [00:05<00:07, 2601514.06it/s]
25%|##4 | 6488064/26421880 [00:05<00:06, 2879025.08it/s]
27%|##7 | 7176192/26421880 [00:06<00:06, 3173470.38it/s]
30%|### | 7929856/26421880 [00:06<00:05, 3481690.43it/s]
33%|###3 | 8749056/26421880 [00:06<00:04, 4100645.60it/s]
35%|###4 | 9207808/26421880 [00:06<00:04, 3928564.95it/s]
39%|###8 | 10190848/26421880 [00:06<00:03, 4427497.65it/s]
43%|####2 | 11239424/26421880 [00:06<00:03, 4864381.85it/s]
47%|####7 | 12419072/26421880 [00:07<00:02, 5380687.53it/s]
52%|#####1 | 13729792/26421880 [00:07<00:02, 5945600.15it/s]
57%|#####7 | 15138816/26421880 [00:07<00:01, 7022425.02it/s]
60%|###### | 15925248/26421880 [00:07<00:01, 6725765.46it/s]
66%|######6 | 17563648/26421880 [00:07<00:01, 8138049.86it/s]
70%|######9 | 18481152/26421880 [00:07<00:01, 7763367.33it/s]
77%|#######7 | 20348928/26421880 [00:08<00:00, 9622700.99it/s]
81%|######## | 21397504/26421880 [00:08<00:00, 8931545.30it/s]
89%|########9 | 23592960/26421880 [00:08<00:00, 10988473.31it/s]
94%|#########3| 24805376/26421880 [00:08<00:00, 10422648.85it/s]
100%|##########| 26421880/26421880 [00:08<00:00, 3095816.12it/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]
100%|##########| 29515/29515 [00:00<00:00, 324595.50it/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]
1%|1 | 65536/4422102 [00:00<00:11, 364151.94it/s]
5%|5 | 229376/4422102 [00:00<00:06, 683437.57it/s]
16%|#6 | 720896/4422102 [00:00<00:01, 2024879.07it/s]
34%|###4 | 1507328/4422102 [00:00<00:00, 3149519.47it/s]
100%|##########| 4422102/4422102 [00:00<00:00, 6090071.00it/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]
100%|##########| 5148/5148 [00:00<00:00, 28636972.14it/s]
Extracting data/FashionMNIST/raw/t10k-labels-idx1-ubyte.gz to data/FashionMNIST/raw
我们将 Dataset 作为 DataLoader 对象传递，包装数据集，并提供了批处理、采样、混洗和多进程数据加载的功能。这里定义了批大小为 64。这样，每次迭代时，数据加载器返回 64 个特征和标签。
batch_size = 64
# 创建数据 loader
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
Shape of X [N, C, H, W]: torch.Size([64, 1, 28, 28])
Shape of y: torch.Size([64]) torch.int64
torch.Size 是一个表示张量（Tensor）维度的元组。
了解更多关于数据加载的内容。
创建模型
定义神经网络需要继承 nn.Module 并实现 forward 方法。我们可以在 __init__ 方法中初始化网络层。
为了加速运算操作，在硬件设备允许的情况系啊，我们可以将其移到 GPU 或者 MPS。
# 获取 cpu， gpu 或 mps 设备以进行训练。
device = (
"cuda"
if torch.cuda.is_available()
else "mps"
if torch.backends.mps.is_available()
else "cpu"
)
print(f"Using {device} device")
# 定义模型类
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)
Using cuda 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)
)
)
了解更多关于创建神经网络的内容。
优化模型参数
要训练模型，我们需要一个损失函数（loss function）和一个优化器（optimizer）。
loss_fn = nn.CrossEntropyLoss()
optimizer = torch.optim.SGD(model.parameters(), lr=1e-3)
在一个训练循环中，模型对训练数据集（以批次形式提供）进行预测，并通过反向传播预测误差来调整模型的参数。
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)
# 计算预测误差
pred = model(X)
loss = loss_fn(pred, y)
# 反向传播 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}]")
我们还需检查模型在测试数据集上的表现，以确保它有成功学习。
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")
训练过程在多个迭代（epoch）中进行。在每个迭代中，模型学习参数以进行更好的预测。我们打印每个迭代的模型准确度（accuracy）和损失（loss）；我们希望看到准确度增加和损失减少。
The training process is conducted over several iterations (epochs). During each epoch, the model learns parameters to make better predictions. We print the model’s accuracy and loss at each epoch; we’d like to see the accuracy increase and the loss decrease with every epoch.
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: 2.303494 [ 64/60000]
loss: 2.294637 [ 6464/60000]
loss: 2.277102 [12864/60000]
loss: 2.269977 [19264/60000]
loss: 2.254235 [25664/60000]
loss: 2.237146 [32064/60000]
loss: 2.231055 [38464/60000]
loss: 2.205037 [44864/60000]
loss: 2.203240 [51264/60000]
loss: 2.170889 [57664/60000]
Test Error:
Accuracy: 53.9%, Avg loss: 2.168588
Epoch 2
-------------------------------
loss: 2.177787 [ 64/60000]
loss: 2.168083 [ 6464/60000]
loss: 2.114910 [12864/60000]
loss: 2.130412 [19264/60000]
loss: 2.087473 [25664/60000]
loss: 2.039670 [32064/60000]
loss: 2.054274 [38464/60000]
loss: 1.985457 [44864/60000]
loss: 1.996023 [51264/60000]
loss: 1.917241 [57664/60000]
Test Error:
Accuracy: 60.2%, Avg loss: 1.920374
Epoch 3
-------------------------------
loss: 1.951705 [ 64/60000]
loss: 1.919516 [ 6464/60000]
loss: 1.808730 [12864/60000]
loss: 1.846550 [19264/60000]
loss: 1.740618 [25664/60000]
loss: 1.698733 [32064/60000]
loss: 1.708889 [38464/60000]
loss: 1.614436 [44864/60000]
loss: 1.646475 [51264/60000]
loss: 1.524308 [57664/60000]
Test Error:
Accuracy: 61.4%, Avg loss: 1.547092
Epoch 4
-------------------------------
loss: 1.612695 [ 64/60000]
loss: 1.570870 [ 6464/60000]
loss: 1.424730 [12864/60000]
loss: 1.489542 [19264/60000]
loss: 1.367256 [25664/60000]
loss: 1.373464 [32064/60000]
loss: 1.376744 [38464/60000]
loss: 1.304962 [44864/60000]
loss: 1.347154 [51264/60000]
loss: 1.230661 [57664/60000]
Test Error:
Accuracy: 62.7%, Avg loss: 1.260891
Epoch 5
-------------------------------
loss: 1.337803 [ 64/60000]
loss: 1.313278 [ 6464/60000]
loss: 1.151837 [12864/60000]
loss: 1.252142 [19264/60000]
loss: 1.123048 [25664/60000]
loss: 1.159531 [32064/60000]
loss: 1.175011 [38464/60000]
loss: 1.115554 [44864/60000]
loss: 1.160974 [51264/60000]
loss: 1.062730 [57664/60000]
Test Error:
Accuracy: 64.6%, Avg loss: 1.087374
Done!
阅读更多关于训练模型的信息。
保存模型
保存模型的一种常见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")
Saved PyTorch Model State to model.pth
加载模型
加载模型的过程包括重新创建模型结构，并将状态字典加载到其中。
model = NeuralNetwork().to(device)
model.load_state_dict(torch.load("model.pth"))
<All keys matched successfully>
这个模型现在可以用来进行预测。
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}"')
Predicted: "Ankle boot", Actual: "Ankle boot"
阅读更多关于保存和加载模型的信息。