刘二大人《PyTorch深度学习实践》完结合集——第11课：卷积神经网络（高级篇）

import torch

import torch
from torchvision import transforms                #针对图像进行处理
from  torchvision import datasets
from torch.utils.data import  DataLoader          #loader picture
import torch.nn.functional as F                   #using function rule
import torch.optim as optim
import matplotlib.pyplot as plt
# 糖尿病预测研判..................................................................................................#
#loader dataset and transform dataset as Tensore
batch_size = 64

transform = transforms.Compose([transforms.ToTensor(),transforms.Normalize((0.1307, ),(0.3081, ))])
#mnist库的训练集当中有60000张图片
train_dataset = datasets.MNIST(root='../dataset/mnist/',train=True,download=True,transform=transform)
train_loader =  DataLoader(train_dataset,shuffle=True,batch_size=batch_size )
#mnist库的训练集当中有10000张图片
test_dataset = datasets.MNIST(root='../dataset/mnist/',train=False,download=True,transform=transform)
test_loader = DataLoader(test_dataset,shuffle=False,batch_size=batch_size )

class Inception(torch.nn.Module):                               #Inception块:输入：[b,in_channels,w,h]   输出：[b,88,w,h]
    def __init__(self,in_channels):
        super(Inception,self).__init__()
        self.branch_pool = torch.nn.Conv2d(in_channels=in_channels,out_channels=24,kernel_size=1)

        self.branch1x1 = torch.nn.Conv2d(in_channels=in_channels,out_channels=16,kernel_size=1)

        self.branch5x5_1 = torch.nn.Conv2d(in_channels=in_channels,out_channels=16,kernel_size=1)
        self.branch5x5_2 = torch.nn.Conv2d(in_channels=16,out_channels=24,kernel_size=5,padding=2)

        self.branch3x3_1 = torch.nn.Conv2d(in_channels=in_channels,out_channels=16,kernel_size=1)
        self.branch3x3_2 = torch.nn.Conv2d(in_channels=16, out_channels=24, kernel_size=3,padding=1)
        self.branch3x3_3 = torch.nn.Conv2d(in_channels=24, out_channels=24, kernel_size=3,padding=1)

    def forward(self, x):
        branch_pool = F.avg_pool2d(x,kernel_size=3,padding=1,stride=1)
        branch_pool = self.branch_pool(branch_pool)   #>>[b,24,w,h]

        branch1x1 = self.branch1x1(x)                 #>>[b,16,w,h]

        branch5x5 = self.branch5x5_1(x)
        branch5x5 = self.branch5x5_2(branch5x5)       #>>[b,24,w,h]

        branch3x3 = self.branch3x3_1(x)
        branch3x3 = self.branch3x3_2(branch3x3)
        branch3x3 = self.branch3x3_3(branch3x3)       #>>[b,24,w,h]

        output = [branch1x1,branch5x5,branch3x3,branch_pool]  #>>[b,16,w,h]+[b,24,w,h]+[b,24,w,h]+[b,24,w,h]
        return torch.cat(output,dim=1)    # >>[b,88,w,h]   仅改变通道数


#Design model and use class.................. inherit from nn.Moduel........................................................#
class  Model(torch.nn.Module):
    def __init__(self):
        super( Model, self).__init__()
        self.conv1 = torch.nn.Conv2d(1,10,kernel_size=5)
        self.conv2 = torch.nn.Conv2d(88,20,kernel_size=5)

        self.inception1 = Inception(in_channels=10)
        self.inception2 = Inception(in_channels=20)

        self.mp = torch.nn.MaxPool2d(2)
        self.fc = torch.nn.Linear(1408,10)
    def forward(self,x):
        x_size = x.size(0)                #x=[b,1,28,28]        x_size = batch_size
        x = F.relu(self.mp(self.conv1(x))) #[b,1,28,28]>[b,10,24,24]>[b,10,12,12]
        x = self.inception1(x)             #[b,88,12,12]
        x = F.relu((self.mp(self.conv2(x)))) #[b,88,12,12]>[b,20,8,8]>[b,20,4,4]
        x = self.inception2(x)              #[b,88,4,4]
        x = x.view(x_size,-1)                #[b,0,1,1408]
        x = self.fc(x)                      #[b,0,1,10]
        return x

model =  Model()#实例化
#Training on GPU........................................................................................................#
device = torch.device("cuda:0"if torch.cuda.is_available() else "cuda:0")
model.to(device)              #model moving on GPU

#Construct loss functions and optimizer.................Use Torch API...................................................#
criterion  =  torch.nn.CrossEntropyLoss()     #老师使用的是torch.nn.BCELoss(size_average=False)但是我使用这个损失太大了
optimizer =   optim.SGD(model.parameters(),lr=0.01,momentum=0.5) #lr为学习率，因为0.01太小了，我改成了0.1


#defining one training..........................................................................................................#
def train(epoch):
    running_loss =0.0
    for batch_idx ,data in enumerate(train_loader,0):   #enumerate函数将train_loader中的元组(mini_batch(四维张量)+target(一维张量))列表化，index从0开始
        inputs ,target =data
        inputs,target = inputs.to(device),target.to(device) #send inputs ,traget at every step to the GPU
        optimizer.zero_grad()

        #forward+backward+update
        outputs = model(inputs)
        loss = criterion(outputs,target)
        loss.backward()
        optimizer.step()

        running_loss +=loss.item()     #不用item，pytorch会构建计算图
        if batch_idx % 300 ==299:
            print('[%d,%5d] loss:%.3f' %(epoch+1,batch_idx+1,running_loss/300))
            running_loss = 0.0

#plot
x_axis = []
y_axis = []

#defining one testing..........................................................................................................#
def test():
    correct = 0
    total = 0

    with torch.no_grad():
        for data in test_loader:
            images,labels = data
            images, labels = images.to(device), labels.to(device)  # send images, labels at every step to the GPU
            outputs = model(images)
            max,predicted = torch.max(outputs.data,dim=1)
            total = total+labels.size(0)
            correct =correct+(predicted == labels).sum().item()
    print('Accuracy on test set:%d %%' %(100*correct/total))
    y_axis.append((correct/total))

if __name__ == '__main__':
    for epoch in range(10):
        train(epoch)
        test()
        x_axis.append(epoch)

#drawing.....................................................................................................#
plt.figure(figsize=(7, 7), dpi=80)  # 创建画布
plt.plot(x_axis,y_axis, color='b', linestyle='-')  # 绘制折线图，点划线
plt.xlabel('epoch')                                                       #设置图x轴标签
plt.ylabel('Accuracy rate')                                                        #设置图y轴标签
plt.legend(["loss"],title='Accuracy&epoch',loc='upper left',fontsize=15)#设置图列
plt.show()                                                                #显示图