1 初识PyTorch¶
1.1 张量
1.导入pytorch包
In [1]:
import torch
2.创建一个未初始化的5x3张量
In [3]:
x = torch.empty(5, 3)
print(x)
3.创建一个随机初始化的5x3张量
In [4]:
x = torch.rand(5, 3)
print(x)
4.创建一个5x3的0张量，类型为long
In [5]:
x = torch.zeros(5, 3, dtype=torch.long)
print(x)
5.直接从数组创建张量
In [6]:
x = torch.tensor([5.5, 3])
print(x)
6.创建一个5x3的单位张量，类型为double
In [7]:
x = torch.ones(5, 3, dtype=torch.double)
print(x)
7.从已有的张量创建相同维度的新张量，并且重新定义类型为float
In [9]:
x = torch.randn_like(x, dtype=torch.float)
print(x)
8.打印一个张量的维度
In [10]:
print(x.size())
9.将两个张量相加
In [11]:
y = torch.rand(5, 3)
print(x + y)
# 方法二
# print(torch.add(x, y))
# 方法三
#result = torch.empty(5, 3)
#torch.add(x, y, out=result)
#print(result)
# 方法四
#y.add_(x)
#print(y)
10.打印张量的第一列
In [12]:
print(x[:, 1])
11.将一个4x4的张量resize成一个一维张量
In [5]:
x = torch.randn(4, 4)
y = x.view(16)
print(x.size(),y.size())
12.将一个4x4的张量，resize成一个2x8的张量
In [ ]:
y = x.view(2, 8)
print(x.size(),y.size())
# 方法二
z = x.view(-1, 8) # 确定一个维度，-1的维度会被自动计算
print(x.size(),z.size())
13.从张量中取出数字
In [6]:
x = torch.randn(1)
print(x)
print(x.item())
1.2 Numpy的操作
14.将张量装换成numpy数组
In [8]:
a = torch.ones(5)
print(a)
b = a.numpy()
print(b)
15.将张量+1，并观察上题中numpy数组的变化
In [9]:
a.add_(1)
print(a)
print(b)
16.从numpy数组创建张量
In [11]:
import numpy as np
a = np.ones(5)
b = torch.from_numpy(a)
print(a)
print(b)
17.将numpy数组+1并观察上题中张量的变化
In [12]:
np.add(a, 1, out=a)
print(a)
print(b)
2 自动微分
2.1 张量的自动微分
18.新建一个张量，并设置requires_grad=True
In [22]:
x = torch.ones(2, 2, requires_grad=True)
print(x)
tensor([[1., 1.],
[1., 1.]], requires_grad=True)
19.对张量进行任意操作（y = x + 2）
In [4]:
y = x + 2
print(y)
print(y.grad_fn) # y就多了一个AddBackward的对象
20.再对y进行任意操作
In [5]:
z = y * y * 3
out = z.mean()
print(z) # z多了MulBackward的对象
print(out) # out多了MeanBackward的对象
2.2 梯度
21.对out进行反向传播
In [9]:
out.backward()
22.打印梯度d(out)/dx
In [10]:
print(x.grad) #out=0.25*Σ3(x+2)^2
tensor([[4.5000, 4.5000],
[4.5000, 4.5000]])
23.创建一个结果为矢量的计算过程（y=x*2^n）
In [11]:
x = torch.randn(3, requires_grad=True)
y = x * 2
while y.data.norm() < 1000:
y = y * 2
print(y)
tensor([-1178.7739, 1015.1417, 861.7645], grad_fn=<MulBackward0>)
24.计算v = [0.1, 1.0, 0.0001]处的梯度
In [12]:
v = torch.tensor([0.1, 1.0, 0.0001], dtype=torch.float)
y.backward(v)
print(x.grad)
tensor([1.0240e+02, 1.0240e+03, 1.0240e-01])
25.关闭梯度的功能
In [13]:
print(x.requires_grad)
print((x ** 2).requires_grad)
with torch.no_grad():
print((x ** 2).requires_grad)
# 方法二
# print(x.requires_grad)
# y = x.detach()
# print(y.requires_grad)
# print(x.eq(y).all())
True
True
False
3 神经网络
这部分会实现LeNet5，结构如下所示
3.1 定义网络
In [2]:
import torch
import torch.nn as nn
import torch.nn.functional as F
class Net(nn.Module):
def __init__(self):
super(Net, self).__init__()
# 26.定义①的卷积层，输入为32x32的图像，卷积核大小5x5卷积核种类6
self.conv1 = nn.Conv2d(3, 6, 5)
# 27.定义③的卷积层，输入为前一层6个特征，卷积核大小5x5，卷积核种类16
self.conv2 = nn.Conv2d(6, 16, 5)
# 28.定义⑤的全链接层，输入为16*5*5，输出为120
self.fc1 = nn.Linear(16 * 5 * 5, 120) # 6*6 from image dimension
# 29.定义⑥的全连接层，输入为120，输出为84
self.fc2 = nn.Linear(120, 84)
# 30.定义⑥的全连接层，输入为84，输出为10
self.fc3 = nn.Linear(84, 10)
def forward(self, x):
# 31.完成input-S2，先卷积+relu，再2x2下采样
x = F.max_pool2d(F.relu(self.conv1(x)), (2, 2))
# 32.完成S2-S4，先卷积+relu，再2x2下采样
x = F.max_pool2d(F.relu(self.conv2(x)), 2) #卷积核方形时，可以只写一个维度
# 33.将特征向量扁平成行向量
x = x.view(-1, 16 * 5 * 5)
# 34.使用fc1+relu
x = F.relu(self.fc1(x))
# 35.使用fc2+relu
x = F.relu(self.fc2(x))
# 36.使用fc3
x = self.fc3(x)
return x
net = Net()
print(net)
Net(
(conv1): Conv2d(3, 6, kernel_size=(5, 5), stride=(1, 1))
(conv2): Conv2d(6, 16, kernel_size=(5, 5), stride=(1, 1))
(fc1): Linear(in_features=400, out_features=120, bias=True)
(fc2): Linear(in_features=120, out_features=84, bias=True)
(fc3): Linear(in_features=84, out_features=10, bias=True)
)
37.打印网络的参数
In [15]:
params = list(net.parameters())
# print(params)
print(len(params))
10
38.打印某一层参数的形状
In [16]:
print(params[0].size())
torch.Size([6, 1, 3, 3])
39.随机输入一个向量，查看前向传播输出
In [27]:
input = torch.randn(1, 3, 32, 32)
out = net(input)
print(out)
tensor([[-0.0672, 0.0347, 0.0879, -0.0769, -0.0172, 0.0181, -0.1240, -0.0204,
0.0145, 0.0922]], grad_fn=<AddmmBackward>)
40.将梯度清零
In [25]:
net.zero_grad()
41.随机一个梯度进行反向传播
In [28]:
out.backward(torch.randn(1, 10))
3.2 损失函数
42.用自带的MSELoss()定义损失函数
In [29]:
criterion = nn.MSELoss()
43.随机一个真值，并用随机的输入计算损失
In [32]:
target = torch.randn(10) # 随机真值
target = target.view(1, -1) # 变成行向量
output = net(input) # 用随机输入计算输出
loss = criterion(output, target) # 计算损失
print(loss)
tensor(1.2790, grad_fn=<MseLossBackward>)
44.将梯度初始化，计算上一步中loss的反向传播
In [ ]:
net.zero_grad()
print('conv1.bias.grad before backward')
print(net.conv1.bias.grad)
45.计算43中loss的反向传播
In [34]:
loss.backward()
print('conv1.bias.grad after backward')
print(net.conv1.bias.grad)
conv1.bias.grad before backward
tensor([0., 0., 0., 0., 0., 0.])
conv1.bias.grad after backward
tensor([ 0.0110, 0.0172, 0.0062, -0.0053, 0.0075, 0.0127])
3.3 更新权重
46.定义SGD优化器算法，学习率设置为0.01
In [3]:
import torch.optim as optim
optimizer = optim.SGD(net.parameters(), lr=0.01)
47.使用优化器更新权重
In [36]:
optimizer.zero_grad()
output = net(input)
loss = criterion(output, target)
loss.backward()
# 更新权重
optimizer.step()
4 训练一个分类器
4.1 读取CIFAR10数据，做标准化
48.构造一个transform，将三通道(0,1)区间的数据转换成(-1,1)的数据
In [4]:
import torchvision
import torchvision.transforms as transforms
transform = transforms.Compose(
[transforms.ToTensor(),
transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))])
读取数据集，挑几个看看啥样（比较繁琐，随便看看，不做题了）
In [5]:
from torch.utils.data import Dataset, DataLoader
import os
import pickle
batch_size = 4
def load_cifar_batch(filename):
""" load single batch of cifar """
with open(filename, 'rb') as f:
datadict = pickle.load(f, encoding='iso-8859-1')
X = datadict['data']
Y = datadict['labels']
X = X.reshape(10000, 3, 32, 32).transpose(0,2,3,1).astype("uint8")
Y = np.array(Y)
return list(zip(X, Y))
def load_cifar(ROOT):
dataset = []
for b in range(1,6):
f = os.path.join(ROOT, 'data_batch_%d' % (b, ))
batch = load_cifar_batch(f)
dataset.append(batch)
data_train = np.concatenate(dataset)
del batch
data_test = load_cifar_batch(os.path.join(ROOT, 'test_batch'))
return data_train, data_test
class cifar(Dataset):
def __init__(self, root, segmentation='train', transforms=None):
if segmentation == 'train':
self.data = load_cifar(root)[0]
elif segmentation == 'test':
self.data = load_cifar(root)[1]
self.transform = transform
def __getitem__(self, index):
data = self.data[index][0]
if(self.transform):
data = (self.transform(data))
else:
data = (torch.from_numpy(data))
label = self.data[index][1]
return data, label
def __len__(self):
return len(self.data)
In [6]:
trainset = cifar(root = '/home/kesci/input/cifar10', segmentation='train', transforms=transform)
testset = cifar(root = '/home/kesci/input/cifar10', segmentation='test', transforms=transform)
trainloader = torch.utils.data.DataLoader(trainset, batch_size=batch_size,shuffle=True, num_workers=2)
testloader = torch.utils.data.DataLoader(testset, batch_size=batch_size,shuffle=False, num_workers=2)
classes = ('plane', 'car', 'bird', 'cat',
'deer', 'dog', 'frog', 'horse', 'ship', 'truck')
In [8]:
import matplotlib.pyplot as plt
import numpy as np
def imshow(img):
img = img / 2 + 0.5 # 把数据退回(0,1)区间
npimg = img.numpy()
plt.imshow(np.transpose(npimg, (1, 2, 0)))
plt.show()
# 随机取一些数据
dataiter = iter(trainloader)
images, labels = dataiter.next()
# 展示图片
imshow(torchvision.utils.make_grid(images))
# 展示分类
print(' '.join('%5s' % classes[labels[j]] for j in range(4)))
frog dog frog deer
4.2 建立网络
这部分沿用前面的网络
In [9]:
net2 = Net()
4.3 定义损失函数和优化器
49.定义交叉熵损失函数
In [10]:
criterion2 = nn.CrossEntropyLoss()
50.定义SGD优化器算法，学习率设置为0.001，momentum=0.9
In [13]:
optimizer2 = optim.SGD(net2.parameters(), lr=0.001, momentum=0.9)
4.4训练网络
In [14]:
for epoch in range(2):
running_loss = 0.0
for i, data in enumerate(trainloader, 0):
# 获取X,y对
inputs, labels = data
# 51.初始化梯度
optimizer2.zero_grad()
# 52.前馈
outputs = net2(inputs)
# 53.计算损失
loss = criterion2(outputs, labels)
# 54.计算梯度
loss.backward()
# 55.更新权值
optimizer2.step()
# 每2000个数据打印平均代价函数值
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')
[1, 2000] loss: 2.178
[1, 4000] loss: 1.835
[1, 6000] loss: 1.678
[1, 8000] loss: 1.585
[1, 10000] loss: 1.537
[1, 12000] loss: 1.466
[2, 2000] loss: 1.423
[2, 4000] loss: 1.371
[2, 6000] loss: 1.363
[2, 8000] loss: 1.351
[2, 10000] loss: 1.317
[2, 12000] loss: 1.327
Finished Training
4.5 使用模型预测
取一些数据
In [17]:
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)))
GroundTruth: cat ship ship plane
56.使用模型预测
In [18]:
outputs = net2(images)
_, predicted = torch.max(outputs, 1)
print('Predicted: ', ' '.join('%5s' % classes[predicted[j]]
for j in range(4)))
Predicted: cat ship ship ship
57.在测试集上进行打分
In [19]:
correct = 0
total = 0
with torch.no_grad():
for data in testloader:
images, labels = data
outputs = net2(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))
Accuracy of the network on the 10000 test images: 52 %
4.6 存取模型
58.保存训练好的模型
In [21]:
PATH = './cifar_net.pth'
torch.save(net.state_dict(), PATH)
59.读取保存的模型
In [25]:
pretrained_net = torch.load(PATH)
60.加载模型
In [27]:
net3 = Net()
net3.load_state_dict(pretrained_net)
Out[27]:
<All keys matched successfully>
