学习pytorch（下）

学习pytorch （下）

——莫烦python之CNN

我通过莫烦python教学视频入门pytorch，通过分析一篇cnn代码样例进行学习。

1. 全局代码

import torch
import torch.nn as nn
from torch.autograd import Variable
import torch.utils.data as Data 
import torchvision
import matplotlib.pyplot as plt 

#Hyper parameter
EPOCH = 1
BATCH_SIZE = 50
LR = 0.001
DOWNLOAD_MNIST = False

train_data = torchvision.datasets.MNIST(
    root = './mnist',
    train = True,
    transform = torchvision.transform.ToTensor(),
    download = DOWNLOAD_MNIST
)

# #plot one example
# print(train_data.train_data.size())
# print(train_data.train_labels.size())
# plt.imshow(train_data.train_data[0].numpy(),cmap='gray')
# plt.title('%i'%train_data.train_labels[0])
# plt.show()

train_loader = Data.DataLoader(
    dataset = train_data,
    batch_size = BATCH_SIZE,
    shuffle = True,
    num_workers = 2
)

test_data = torchvision.datasets.MNIST(
    root = './mnist',
    train = False
)

test_x = Variable(torch.unsequeeze(test_data.test_data,dim = 1),volatile=True).type(torch.FloatTensor)[:2000]/255
test_y = test_data.test_lables[:2000]

class CNN(nn.Module):
    def __init__(self):
        super(CNN,self).__init__()
        self.conv1 = nn.Sequential(
            nn.Conv2d(
                in_channels = 1,
                out_channels = 16,
                kernerl_size = 5,
                stride = 1,
                padding =2,
            ),
            nn.ReLU(),
            nn.MaxPool2d(kernerl_size = 2),
        )

        self.conv2 = nn.Sequential(
            nn.Conv2d(16,32,5,1,2),
            nn.ReLU(),
            nn.MaxPool2d(2),
        )
        
        self.out = nn.Linear(32*7*7,10)

    def forward(self,x):
        x = self.conv1(x)
        x = self.conv2(x)           # (batch, 32, 7, 7)
        x = x.view(x.size(0),-1)    # (batch, 32 *7 *7)   x.size(0) keep the size of batch
        output = self.out(x)
        return output

cnn = CNN()
print(cnn)

optimizer = torch.optim.Adam(cnn.parameters(),lr=LR)
loss_func = nn.CrossEntropyLoss()

#training and testing
for epoch in range(EPOCH):
    for step,(x,y) in enumerate(train_loader):
        b_x = Variable(x)
        b_y = Variable(y)

        output = cnn(b_x)
        loss = loss_func(output,b_y)
        optimizer.zero_grad()
        loss.backward()
        optimizer.step()

        if step % 50 == 0:
            test_output = cnn(test_x)
            pred_y = torch.max(test_output,1)[1].data.squeeze()
            accuracy = sum(pred_y == test_y)/test_.size(0)
            print('EPOCH:',epoch,'| train_loss:', loss.data[0], '| test accuracy:',accuracy)
        

#print 10 predictions from test data
test_output = cnn(test_x[:10])
pred_y = torch.max(test_output,1)[1].data.numpy().squeeze()
print(pred_y,'redictin number')
print(test_y[:10].numpy(),'real number')

2. 局部分析

加载相应的库

import torch
import torch.nn as nn
from torch.autograd import Variable
import torch.utils.data as Data 
import torchvision
import matplotlib.pyplot as plt 

超参数

EPOCH = 1
BATCH_SIZE = 50
LR = 0.001
DOWNLOAD_MNIST = False

EPOCH：遍历几次数据集
BATCH_SIZE: 批训练大小
LR：学习率
DOWNLOAD_MNIST: 标志位（是否下载数据集，第一次置True，之后置False）

通过torchvision下载数据集MNIST

train_data = torchvision.datasets.MNIST(
    root = './mnist', #数据集存放目录
    train = True, #取训练数据部分
    transform = torchvision.transform.ToTensor(),#转至Tensor格式
    download = DOWNLOAD_MNIST #标志位，是否下载
)

通过Loader准备好训练数据

train_loader = Data.DataLoader(  #Data.DataLoader是必须步骤
    dataset = train_data, #加载数据
    batch_size = BATCH_SIZE, #批训练
    shuffle = True, #标志位，是否打乱数据集顺序
    num_workers = 2 #开几个线程
)

准备测试数据集

test_data = torchvision.datasets.MNIST(
    root = './mnist',
    train = False
)

选择前2000个数据，将训练数据变成Variable

test_x = Variable(torch.unsequeeze(test_data.test_data,dim = 1),volatile=True).type(torch.FloatTensor)[:2000]/255
test_y = test_data.test_lables[:2000]

搭建网络结构

class CNN(nn.Module):
    def __init__(self):
        super(CNN,self).__init__()
        self.conv1 = nn.Sequential( #卷积神经网络
            nn.Conv2d(
                in_channels = 1, #输入通道
                out_channels = 16,#输出通道
                kernerl_size = 5,#核的大小
                stride = 1,
                padding =2,
            ),
            nn.ReLU(),
            nn.MaxPool2d(kernerl_size = 2),#池化，2*2
        )

        self.conv2 = nn.Sequential(
            nn.Conv2d(16,32,5,1,2),
            nn.ReLU(),
            nn.MaxPool2d(2),
        )
        
        self.out = nn.Linear(32*7*7,10)

    def forward(self,x):  #定义数据传输
        x = self.conv1(x)
        x = self.conv2(x)           # (batch, 32, 7, 7)
        x = x.view(x.size(0),-1)    # (batch, 32 *7 *7)   x.size(0) keep the size of batch
        output = self.out(x)
        return output

定义优化器核损失函数

optimizer = torch.optim.Adam(cnn.parameters(),lr=LR)
loss_func = nn.CrossEntropyLoss()

训练过程

for epoch in range(EPOCH):
    for step,(x,y) in enumerate(train_loader):
        b_x = Variable(x) #将训练数据里面的特征变成Variable
        b_y = Variable(y) #将训练数据里面的标签变成Variable

        output = cnn(b_x) #预测结果
        loss = loss_func(output,b_y) #损失
        optimizer.zero_grad() #每次将优化器的导数先置为0
        loss.backward() #后向传播
        optimizer.step() #优化

        if step % 50 == 0: #每隔50步测试一下网络效果
            test_output = cnn(test_x) #输入测试数据，Variable形式
            pred_y = torch.max(test_output,1)[1].data.squeeze()
            accuracy = sum(pred_y == test_y)/test_.size(0)
            print('EPOCH:',epoch,'| train_loss:', loss.data[0], '| test accuracy:',accuracy)

3.总结

综上，步骤可以总结为：