2.1 pytorch快速入门
本文主要介绍机器学习中常见任务的API。
处理数据
PyTorch有两个处理数据的方式:torch.utils.data.DataLoader
和torch.utils.data.Dataset
。
Dataset
存储样本及其相应的标签,
DataLoader
在Dataset
的外层用迭代器进行包装。
import torch
from torch import nn
from torch.utils.data import DataLoader
from torchvision import datasets
from torchvision.transforms import ToTensor
PyTorch提供特定于领域的库,如TorchText、TorchVision和TorchAudio,所有这些库都包括数据集。在本教程中,我们将使用TorchVision数据集。
torchvision.datasets
模块包含许多真实世界视觉数据的Dataset
对象,如CIFAR,COCO. 在本教程中,我们使用FashionMNIST数据集。每个TorchVisionDataset
都包括两个参数:transform
和target_transform
以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(),
)
输出是这样的:
D:\testproj\pytorchOne\venvtorch\Scripts\python.exe D:\testproj\pytorchOne\quickstart.py
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
100.0%
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
100.0%
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
100.0%
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
100.0%
Extracting data\FashionMNIST\raw\t10k-labels-idx1-ubyte.gz to data\FashionMNIST\raw
Process finished with exit code 0
在工程的根目录下添加了如下文件:
我们将数据集作为参数传递给DataLoader。这在我们的数据集上封装了一个可迭代的,并支持自动批处理、采样、混洗和多进程数据加载。在这里,我们定义了一个64的批量大小,即数据加载器可迭代中的每个元素都将返回一批(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
我们来看看输出:
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中定义神经网络,我们创建了一个继承自nn.Module的类。我们在__init__函数中定义网络的层,并在前向函数中指定数据如何通过网络。为了加速神经网络中的操作,我们将其移动到GPU或MPS(如果可用)。
# 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)
输出如下:
Using cpu 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)
)
)
优化模型参数
为了训练模型,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(分批提供给它),并对预测误差进行反向传播以调整模型的参数。
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
optimizer.zero_grad()
loss.backward()
optimizer.step()
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")
训练过程是通过几个迭代(epochs)进行的。在每个历时中,模型学习参数以做出更好的预测。我们在每个历时中打印模型的准确度和损失;我们希望看到准确度在每个历时中增加,损失在每个历时中减少。
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!")
查看训练过程的输出
Epoch 1
-------------------------------
loss: 2.298773 [ 64/60000]
loss: 2.288743 [ 6464/60000]
loss: 2.266891 [12864/60000]
loss: 2.269755 [19264/60000]
loss: 2.260278 [25664/60000]
loss: 2.214578 [32064/60000]
loss: 2.232736 [38464/60000]
loss: 2.191824 [44864/60000]
loss: 2.183783 [51264/60000]
loss: 2.164333 [57664/60000]
Test Error:
Accuracy: 38.6%, Avg loss: 2.161599
Epoch 2
-------------------------------
loss: 2.159602 [ 64/60000]
loss: 2.153630 [ 6464/60000]
loss: 2.097003 [12864/60000]
loss: 2.125920 [19264/60000]
loss: 2.080615 [25664/60000]
loss: 1.998236 [32064/60000]
loss: 2.039693 [38464/60000]
loss: 1.950894 [44864/60000]
loss: 1.957268 [51264/60000]
loss: 1.893443 [57664/60000]
Test Error:
Accuracy: 54.2%, Avg loss: 1.897607
Epoch 3
-------------------------------
loss: 1.916339 [ 64/60000]
loss: 1.889156 [ 6464/60000]
loss: 1.775602 [12864/60000]
loss: 1.833217 [19264/60000]
loss: 1.727981 [25664/60000]
loss: 1.651068 [32064/60000]
loss: 1.687254 [38464/60000]
loss: 1.577951 [44864/60000]
loss: 1.611350 [51264/60000]
loss: 1.507689 [57664/60000]
Test Error:
Accuracy: 63.2%, Avg loss: 1.528882
Epoch 4
-------------------------------
loss: 1.583922 [ 64/60000]
loss: 1.547395 [ 6464/60000]
loss: 1.399195 [12864/60000]
loss: 1.486909 [19264/60000]
loss: 1.374229 [25664/60000]
loss: 1.344481 [32064/60000]
loss: 1.368184 [38464/60000]
loss: 1.283136 [44864/60000]
loss: 1.326591 [51264/60000]
loss: 1.228549 [57664/60000]
Test Error:
Accuracy: 64.5%, Avg loss: 1.252939
Epoch 5
-------------------------------
loss: 1.319819 [ 64/60000]
loss: 1.299578 [ 6464/60000]
loss: 1.135219 [12864/60000]
loss: 1.254325 [19264/60000]
loss: 1.136180 [25664/60000]
loss: 1.141785 [32064/60000]
loss: 1.167674 [38464/60000]
loss: 1.097063 [44864/60000]
loss: 1.143751 [51264/60000]
loss: 1.062997 [57664/60000]
Test Error:
Accuracy: 65.1%, Avg loss: 1.080314
Done!
保存模型
保存模型的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")
点击查看保存模型的输出
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()
model.load_state_dict(torch.load("model.pth"))
这个模型现在可以用来进行预测了。
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():
pred = model(x)
predicted, actual = classes[pred[0].argmax(0)], classes[y]
print(f'Predicted: "{predicted}", Actual: "{actual}"')
预测结果的输出是:
Predicted: "Ankle boot", Actual: "Ankle boot"