目录
TRAINING A CLASSIFIER 训练一个分类器
就这样. 你已经知道怎样定义神经网络, 计算loss和更新网络权重.
现在你可能会想:
数据呢?
一般, 当处理图片, 文本, 音频或者视频数据时, 你可以使用标准的python包,将数据载入为numpy array. 然后你可以将array转化为torch.*Tensor
.
对于图片, Pillow, OpenCV之类的包较为有用.
对于音频, 可以用scipy和librosa
对于文本, 基于原生Python或Cython的载入功能, NLTK和SpaCy之类的较为有用.
特别对于视觉(vision), 我们创建了一个包叫torchvision
. 它有常用datasets(如ImageNet, CIFAR10, MNIST等)的数据载入工具(data loader)和图片转换工具, 也就是torchvision.datasets
和torch.utils.data.DataLoader
.
用起来很方便.
这个教程里. 我们用CIFAR10数据库. 有 ‘airplane’, ‘automobile’, ‘bird’, ‘cat’, ‘deer’, ‘dog’, ‘frog’, ‘horse’, ‘ship’, ‘truck’十个类别的图片. 图片3x32x32. 也就是3通道, 32x32大小.
Training an image classifier 训练一个图像分类器
步骤:
- 用
torchvision
载入和归一化训练和测试数据, - 定义卷积神经网络(CNN)
- 定义loss函数
- 训练
- 测试
1. 载入和归一化CIFAR10
使用torchvision
, 很简单.
import torch
import torchvision
import torchvision.transforms as transforms
torchvision
的输出是PILImage, 范围[0,1]. 我们将其转化为归一化到[-1, 1]范围的Tensor
.
注意:
如果在Windows上出现BrokenPipeError
, 将torch.utils.data.DataLoader()
的num_worker
置为0.
transform = transforms.Compose(
[transforms.ToTensor(),
transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))])
trainset = torchvision.datasets.CIFAR10(root='./data', train=True,
download=True, transform=transform)
trainloader = torch.utils.data.DataLoader(trainset, batch_size=4,
shuffle=True, num_workers=2)
testset = torchvision.datasets.CIFAR10(root='./data', train=False,
download=True, transform=transform)
testloader = torch.utils.data.DataLoader(testset, batch_size=4,
shuffle=False, num_workers=2)
classes = ('plane', 'car', 'bird', 'cat',
'deer', 'dog', 'frog', 'horse', 'ship', 'truck')
Out:
Downloading https://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz to ./data/cifar-10-python.tar.gz
Extracting ./data/cifar-10-python.tar.gz to ./data
Files already downloaded and verified
让我们看看训练数据什么样.
import matplotlib.pyplot as plt
import numpy as np
# functions to show an image
def imshow(img):
img = img / 2 + 0.5 # unnormalize
npimg = img.numpy()
plt.imshow(np.transpose(npimg, (1, 2, 0)))
plt.show()
# get some random training images
dataiter = iter(trainloader)
images, labels = dataiter.next()
# show images
imshow(torchvision.utils.make_grid(images))
# print labels
print(' '.join('%5s' % classes[labels[j]] for j in range(4)))
../../_images/sphx_glr_cifar10_tutorial_001.png
Out:
dog bird ship dog
2. Define a Convolutional Neural Network 定义一个卷积神经网络
将之前定义的网络拷贝过来, 修改为接收3通道图片(之前的接收单通道).
import torch.nn as nn
import torch.nn.functional as F
class Net(nn.Module):
def __init__(self):
super(Net, self).__init__()
self.conv1 = nn.Conv2d(3, 6, 5)
self.pool = nn.MaxPool2d(2, 2)
self.conv2 = nn.Conv2d(6, 16, 5)
self.fc1 = nn.Linear(16 * 5 * 5, 120)
self.fc2 = nn.Linear(120, 84)
self.fc3 = nn.Linear(84, 10)
def forward(self, x):
x = self.pool(F.relu(self.conv1(x)))
x = self.pool(F.relu(self.conv2(x)))
x = x.view(-1, 16 * 5 * 5)
x = F.relu(self.fc1(x))
x = F.relu(self.fc2(x))
x = self.fc3(x)
return x
net = Net()
3. Define a Loss function and optimizer 定义一个loss函数和优化器
用分类交叉熵和带动量(momentum)的SGD.
import torch.optim as optim
criterion = nn.CrossEntropyLoss()
optimizer = optim.SGD(net.parameters(), lr=0.001, momentum=0.9)
4. Train the network 训练网络
有意思的地方来了. 我们简单地在数据迭代器(data iterator)上循环遍历, 将输入喂给网络并优化.
for epoch in range(2): # loop over the dataset multiple times
running_loss = 0.0
for i, data in enumerate(trainloader, 0):
# get the inputs; data is a list of [inputs, labels]
inputs, labels = data
# zero the parameter gradients
optimizer.zero_grad()
# forward + backward + optimize
outputs = net(inputs)
loss = criterion(outputs, labels)
loss.backward()
optimizer.step()
# print statistics
running_loss += loss.item()
if i % 2000 == 1999: # print every 2000 mini-batches
print('[%d, %5d] loss: %.3f' %
(epoch + 1, i + 1, running_loss / 2000))
running_loss = 0.0
print('Finished Training')
Out:
[1, 2000] loss: 2.235
[1, 4000] loss: 1.902
[1, 6000] loss: 1.674
[1, 8000] loss: 1.571
[1, 10000] loss: 1.529
[1, 12000] loss: 1.478
[2, 2000] loss: 1.389
[2, 4000] loss: 1.391
[2, 6000] loss: 1.359
[2, 8000] loss: 1.335
[2, 10000] loss: 1.330
[2, 12000] loss: 1.289
Finished Training
保存训练好的模型:
PATH = './cifar_net.pth'
torch.save(net.state_dict(), PATH)
这里有更多的保存模型的细节.
5. Test the network on the test data 用测试数据测试网络
我们已经训练了两轮了, 我们看看网络是否学到东西了.
我们将模型输出的结果和真实结果(ground-truth)作比较. 如果预测是正确的, 我们将这个sample加到正确预测的list中.
dataiter = iter(testloader)
images, labels = dataiter.next()
# print images
imshow(torchvision.utils.make_grid(images))
print('GroundTruth: ', ' '.join('%5s' % classes[labels[j]] for j in range(4)))
Out:
GroundTruth: cat ship ship plane
然后, 重新载入保存的模型(注意: 保存和重新载入模型不是必要的, 我们这里只是为了说明怎样做):
net = Net()
net.load_state_dict(torch.load(PATH))
好, 我们来看看模型分类结果如何:
outputs = net(images)
输出是10类的置信度(energy). 某一类的置信度越高, 网络认为图片是此类的可能性越大. 所以, 我们来获取最高的置信度:
_, predicted = torch.max(outputs, 1)
print('Predicted: ', ' '.join('%5s' % classes[predicted[j]]
for j in range(4)))
Out:
Predicted: cat ship ship ship
结果还不错, 我们来看看整个数据集上的表现:
correct = 0
total = 0
with torch.no_grad():
for data in testloader:
images, labels = data
outputs = net(images)
_, predicted = torch.max(outputs.data, 1)
total += labels.size(0)
correct += (predicted == labels).sum().item()
print('Accuracy of the network on the 10000 test images: %d %%' % (
100 * correct / total))
Out:
Accuracy of the network on the 10000 test images: 53 %
看起来比随机猜测要好. 随机猜测正确率是10%. 看来网络确实学到了东西.
看看哪些类表现好, 哪些类表现不好:
class_correct = list(0. for i in range(10))
class_total = list(0. for i in range(10))
with torch.no_grad():
for data in testloader:
images, labels = data
outputs = net(images)
_, predicted = torch.max(outputs, 1)
c = (predicted == labels).squeeze()
for i in range(4):
label = labels[i]
class_correct[label] += c[i].item()
class_total[label] += 1
for i in range(10):
print('Accuracy of %5s : %2d %%' % (
classes[i], 100 * class_correct[i] / class_total[i]))
Out:
Accuracy of plane : 56 %
Accuracy of car : 74 %
Accuracy of bird : 30 %
Accuracy of cat : 29 %
Accuracy of deer : 57 %
Accuracy of dog : 37 %
Accuracy of frog : 73 %
Accuracy of horse : 62 %
Accuracy of ship : 56 %
Accuracy of truck : 53 %
好了, 接下来做什么呢?
怎样在GPU上跑神经网络呢?
Training on GPU GPU上训练
正如将Tensor
转到GPU上一样, 你可以将网络转到GPU上.
如果CUDA有效的话, 将device定义为第一个可用的cuda设备.
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
# Assuming that we are on a CUDA machine, this should print a CUDA device:
print(device)
Out:
cuda:0
剩下的部分, 我们认为device就是CUDA device.
然后这些方法将会递归地遍历我们的模型部件(module)并将他们的参数和buffer转为CUDA tensor:
net.to(device)
记住, 你必须要在每一步中将输入图像和目标输出都转到GPU上:
inputs, labels = data[0].to(device), data[1].to(device)
为什么我们有看到很大的速度提升? 因为你的模型太小了.
练习: 尝试增大网络宽度(第一个nn.Conv2d
的第二个参数, 第二个nn.Conv2d
的第一个参数, 两者要相等), 看看提速效果如何.
目标达成:
- 在高层(high level)上理解Pytorch Tensor库和神经网络.
- 训练一个分类图片的小神经网络.
Training on multiple GPUs 在多GPU上训练.
如果你想用多块GPU训练, 请看: Data Parallelism.
Where do I go next? 下面做什么?
Total running time of the script: ( 3 minutes 29.428 seconds)