PyTorch初学

初识PyTorch

1.张量

import torch

#创建一个空的5x3张量
x = torch.empty(5, 3)

#创建一个随机初始化的5x3张量
x = torch.rand(5, 3)

# 创建一个5x3的0张量，类型为long
x = torch.zeros(5, 3, dtype=torch.long)

# 创建一个5x3的单位张量，类型为double
x = torch.ones(5, 3, dtype=torch.double)

# 直接从数组创建张量
x = torch.tensor([5.5, 3])

#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
x = torch.randn_like(x, dtype=torch.float)

# 打印一个张量的维度
print(x.size())

# 将两个张量相加
y = torch.rand(5, 3)
print(x + y)
print(torch.add(x, y))
result = torch.zeros(5, 3)
torch.add(x, y, out=result)
print(result)
y.add_(x)
print(y)

# 取张量的第一列
print(x[:, 1])

# 将一个4✖4的张量resize成一个一维张量, 和一个n✖8维的二维向量
x = torch.randn(4, 4)
y = x.view(16)
z = x.view(-1, 8)
print(x.size(), y.size(), z.size())

# 从张量中取出数字
x = torch.randn(1)
print(x)
print(x.item())

2.Numpy的操作

# 将张量转化为Numpy数组
a = torch.ones(5)
b = a.numpy()

# 将张量加1，观察b的变化
a.add_(1)
print(a)
print(b)

# 从numpy数组创建张量
import numpy as np
a = np.ones(5)
b = torch.from_numpy(a)
print(a, b)

# 将numpy数组+1， 观察变化
np.add(a, 1, out=a)
print(a, b)

张量的自动微分

# 新建一个张量，并设置 requires_grad=True
x = torch.ones(2, 2, requires_grad=True)
print(x)

# 对张量进行任意操作（y=x+2）
y = x + 2
print(y)
print(y.grad_fn)# y就多了一个AddBackward

# 再对y进行任意操作
z = y * y * 3
out = z.mean()
print(z)  # z多了MulBackward
print(out) # out 多了MeanBackward

梯度

#对out进行反向传播
out.backward()

# 打印梯度d(out)/dx
print(x.grad)

# 待补充

神经网络

这部分会实现LeNet5，结构如下所示：

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.将特征向量扁平成列向量
        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)

# 打印网络的参数
params = list(net.parameters())
print(params)

# 打印某一层参数的形状
print(params[0].size())

#随机输入一个向量，查看前向传播输出
input = torch.randn(1, 1, 32, 32)
out = net(input)
print(out)

# 将梯度初始化
net.zero_grad()

#随机一个梯度进行反向传播
out.backward(torch.randn(1, 10))

##########损失函数#################
# 用自带的MSELoss()定义损失函数
criterion = nn.MSELoss()

# 随机一个真值，并用随机的输入计算损失
target = torch.randn(10) #随机真值
target = torch.view(1, -1)  #变成列向量

output = net(input) #用随机输入计算输出

loss = criterion(output, target)  #计算损失
print(loss)

# 将梯度初始化，计算上一步中loss的反向传播
net.zero_gard()

print('conv1.bias.grad before backward')
print(net.conv1.bias.grad)

loss.backward()

print('conv1.bias.grad after backward')
print(net.conv1.bias.grad)

##########################更新权重
# 定义SGD优化器算法，学习率设置为0.01
import torch.optim as optim
optimizer = optim.SGD(net.parameters(), lr=0.01)

# 使用优化器更新权重
optimizer.zero_grad()
output = net(input)
loss = criterion(output, target)
loss.backward()

# 更新权重
optimizer.step()

训练一个分类器

# 读取CIFAR10数据，做标准化
# 构造一个transform, 将三通道(0, 1)区间的数据转换成(-1, 1)的数据
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
trainset = cifar(root = './input/cifar10', segmentation='train', transforms=transform)
testset = cifar(root = './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')

# 建立网络
net2 = Net()
# 定义损失函数和优化器
#定义交叉熵损失函数
criterion2 = nn.CrossEntropyLoss()

#定义SGD优化器算法，学习率置为0.001， momentum=0.9
optimizer2 = optim.SGD(net2.parameters(), lr=0.001, momentum=0.9)

#训练网络
for epoch in range(2):
	running_loss = 0.0
	for i, data in enumerate(trainloader, 0):
		# 获取x，y对
		inputs, labels = data
		#初始化梯度
		optimizer2.zero_grad()
		#前馈
		outputs = net2(inputs)
		#计算损失
		loss = criterion2(outputs, labels)
		#计算梯度
		loss.backward()
		#更新权值
		optimizer2.step()
		#每2000个数据打印平均代价函数
		running_loss += loss.item()
		if i%2000 == 1999:
			print('[%d, %5d]  loss: %.3f' %(epoch+1, i+1, running_loss/2000))
			running_loss = 0.0
print('Finished Training')
		 

使用模型预测

# 取一些数据
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)))

# 使用模型预测
outputs = net2(images)
_, predicted = torch.max(outputs, 1)
print('Predicted: ', ' '.join('%5s' % classes[predicted[j]]
                              for j in range(4)))

#在测试集上进行打分
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))

存取模型

# 保存训练好的模型
PATH = './cifar_net.pth'
torch.save(net.state_dict(), PATH)

# 读取保存的模型
pretrained_net = torch.load(PATH)

# 加载模型
net3 = Net()
net3.load_state_dict(pretrained_net)