第一种方法

1.读取数据

import torch 
from torchvision import transforms
from torchvision import datasets
from torch.utils.data import DataLoader
import torch.nn.functional as F
import torch.optim as optim
import random
import matplotlib.pyplot as plt

batchSize=64
transform=transforms.Compose([
    transforms.ToTensor(),
    #均值=0.1307，标准差=0.3081
    transforms.Normalize((0.1307,),(0.3081,))
])
train_dataset=datasets.MNIST(root="data/mnist",
                          train=True,
                          download=True,
                          transform=transform)
train_loader=DataLoader(train_dataset,
                        shuffle=True,
                        batch_size=batchSize)

test_dataset=datasets.MNIST(root="data/mnist",
                          train=False,
                          download=True,
                          transform=transform)
test_loader=DataLoader(test_dataset,
                        shuffle=True,
                        batch_size=batchSize)

2.随机显示一个数字

def showRandomPictures(train_loader):
    data_iter=iter(train_loader)
    images,labels=data_iter.__next__()
    # Choose a random index from the batch
    random_index = random.randint(0, batchSize - 1)

    # Display the transformed image
    transformed_image = images[random_index].squeeze().numpy()
    transformed_image = (transformed_image * 0.3081) + 0.1307  # Inverse normalization
    plt.imshow(transformed_image, cmap='gray')  # Assuming MNIST images are grayscale
    plt.title(f"Label: {labels[random_index].item()}")
    plt.show() 
showRandomPictures(train_loader)

 

3.定义模型和超参数

class Net(torch.nn.Module):
    def __init__(self) -> None:
        super(Net,self).__init__()
        self.layer1=torch.nn.Linear(784,512)
        self.layer2=torch.nn.Linear(512,256)
        self.layer3=torch.nn.Linear(256,128)
        self.layer4=torch.nn.Linear(128,64)
        self.layer5=torch.nn.Linear(64,10)
    def forward(self,x):
        x=x.view(-1,784)
        x=F.relu(self.layer1(x))
        x=F.relu(self.layer2(x))
        x=F.relu(self.layer3(x))
        x=F.relu(self.layer4(x))
        x=self.layer5(x)
        return x
    
model=Net()
criterion=torch.nn.CrossEntropyLoss()
optimizer=optim.SGD(model.parameters(),lr=0.01,momentum=0.5)

4.训练步骤

def train(epoch):
    running_loss=0.0
    for batch_idx,data in enumerate(train_loader,0):
        inputs,target=data
        optimizer.zero_grad()

        outputs=model(inputs)
        loss=criterion(outputs,target)
        loss.backward()
        optimizer.step()

        running_loss+=loss.item()
        if batch_idx%300==299:
            print(f"[epoch={epoch+1}, batch_index={batch_idx+1}, training_loss={running_loss/300}]")
            running_loss=0.0

5.测试步骤 

accuracyHistory=[]
def test():
    correct=0
    total=0
    with torch.no_grad():
        for data in test_loader:
            images,labels=data
            outputs=model(images)
            #每一行的最大值的下标[max,maxIndex]
            _,predicted=torch.max(outputs.data,dim=1)
            #label.size(0)是batch_size
            total+=labels.size(0)
            correct+=(predicted==labels).sum().item()
    print(f"Accuracy on test set:{100*correct/total}")
    accuracyHistory.append(100*correct/total)

6.主方法执行

if __name__=="__main__":
    for epoch in range(10):
        train(epoch)
        test()
    plt.plot(accuracyHistory)
    plt.show()

 7.运行结果

[epoch=1, batch_index=300, training_loss=2.223298035860062]
[epoch=1, batch_index=600, training_loss=0.9641364443302155]
[epoch=1, batch_index=900, training_loss=0.43318099692463874]
Accuracy on test set:88.38
....
Accuracy on test set:97.34
[epoch=8, batch_index=300, training_loss=0.05266832955181599]
[epoch=8, batch_index=600, training_loss=0.05094206769950688]
[epoch=8, batch_index=900, training_loss=0.053400517796787124]
Accuracy on test set:97.49
[epoch=9, batch_index=300, training_loss=0.036180642331019044]
[epoch=9, batch_index=600, training_loss=0.04376721960259602]
[epoch=9, batch_index=900, training_loss=0.04324531762007003]
Accuracy on test set:97.55
[epoch=10, batch_index=300, training_loss=0.02999880815623328]
[epoch=10, batch_index=600, training_loss=0.0335875346181759]
[epoch=10, batch_index=900, training_loss=0.03751158087824782]
Accuracy on test set:97.61

第二种方法

1.下载mnist.pkl.gz 

网址：http://www.iro.umontreal.ca/~lisa/deep/data/mnist/mnist.pkl.gz

数据集文件夹路径是data2/mnist/mnist.pkl.gz

 2.读取数据

from pathlib import Path

import matplotlib.pyplot as plt

DATA_PATH=Path("./data2")
PATH=DATA_PATH / "mnist"
FILENAME="mnist.pkl.gz"
import pickle
import gzip
with gzip.open((PATH/FILENAME).as_posix(),"rb") as f:
    ((x_train,y_train),(x_valid,y_valid),_)=pickle.load(f,encoding="latin-1")
#x_train(500,784),y_train(5000,) x_valid(10000, 784),y_valid(10000,)

随机显示一个数字

#==========28*28=784========随机显示数字5
import matplotlib.pyplot as plt
import numpy as np
plt.imshow(x_train[50].reshape((28,28)),cmap="gray")
plt.show()

 

数据转为tensor

#=================数据转为tensor才能参与建模训练===
import torch
x_train,y_train,x_valid,y_valid=map(
    torch.tensor, (x_train,y_train,x_valid,y_valid)
)

3.设置损失函数为交叉熵函数

#=====torch.nn.functional==========
import torch.nn.functional as F
loss_func=F.cross_entropy

4.创建Model类

from torch import nn
class Mnist_NN(nn.Module):
    def __init__(self):
        super().__init__()
        self.hidden1=nn.Linear(784,128)
        self.hidden2=nn.Linear(128,256)
        self.out=nn.Linear(256,10)
        self.dropout=nn.Dropout(0.5)
    def forward(self,x):
        x=F.relu(self.hidden1(x))
        #全连接层+droput，防止过拟合
        x=self.dropout(x)

        x=F.relu(self.hidden2(x))
        x=self.dropout(x)

        x=self.out(x)
        return x

# Mnist_NN(
#   (hidden1): Linear(in_features=784, out_features=128, bias=True)
#   (hidden2): Linear(in_features=128, out_features=256, bias=True)
#   (out): Linear(in_features=256, out_features=10, bias=True)
#   (dropout): Dropout(p=0.5, inplace=False)
# )
# net=Mnist_NN()
# print(net)

打印一下这网络长什么样

net=Mnist_NN()
print(net)

#打印定义好的名字和w和b
for name,parameter in net.named_parameters():
    print(name,parameter,parameter.size())

Mnist_NN(
  (hidden1): Linear(in_features=784, out_features=128, bias=True)
  (hidden2): Linear(in_features=128, out_features=256, bias=True)
  (out): Linear(in_features=256, out_features=10, bias=True)
  (dropout): Dropout(p=0.5, inplace=False)
)
hidden1.weight Parameter containing:
tensor([[-1.7000e-02, -7.5721e-03, -1.7358e-03,  ...,  7.6538e-03,
..........
out.weight Parameter containing:
tensor([[-0.0173,  0.0522,  0.0494,  ..., -0.0579, -0.0439, -0.0522],
       ....
       requires_grad=True) torch.Size([10, 256])
out.bias Parameter containing:
tensor([-0.0154, -0.0028, -0.0574, -0.0608, -0.0276,  0.0483,  0.0503,  0.0112,
        -0.0352, -0.0498], requires_grad=True) torch.Size([10])

5.使用TensorDataset和DataLoader，封装成一个batch的数据集

from torch.utils.data import TensorDataset
from torch.utils.data import DataLoader
bs=64
train_ds=TensorDataset(x_train,y_train)
# train_dl=DataLoader(train_ds,batch_size=bs,shuffle=True)

valid_ds=TensorDataset(x_valid,y_valid)
# valid_dl=DataLoader(valid_ds,batch_size=bs*2)

def get_data(train_ds,valid_ds,bs):
    return (
        DataLoader(train_ds,batch_size=bs,shuffle=True),
        DataLoader(valid_ds,batch_size=bs*2)
    )

6.定义训练步骤

import numpy as np
val_losses=[]
#steps:迭代次数,step相当于epoch
def fit(steps,model,loss_func,opt,train_dl,valid_dl):
    for step in range(steps):
        model.train() #更新w和b
        #xb(64,784) yb(64),xb和yb都是tensor
        for xb,yb in train_dl:
            loss_batch(model,loss_func,xb,yb,opt)
        #evaluate 模式，dropout和BatchNum不会工作
        model.eval() #不更新w和b
        with torch.no_grad():
            #losses：nums=(loss,batch),(loss,batch)....
            losses,nums =zip(
                *[loss_batch(model,loss_func,xb,yb) for xb,yb in valid_dl]
            )

        #总的验证集的平均损失
        val_loss=np.sum(np.multiply(losses,nums)) / np.sum(nums)
        val_losses.append(val_loss)
        print("当前step:"+str(step),"验证集损失"+str(val_loss))

from torch import optim
def get_model():
    model=Mnist_NN()
    #返回模型和优化器optim.SGD(model.parameters() , lr=0.001)
    return model,optim.Adam(model.parameters() , lr=0.001)


def loss_batch(model, loss_func ,xb,yb, opt=None):
    #根据预测值和真实值计算loss
    loss=loss_func( model(xb) , yb )

    if opt is not None:
        loss.backward()  #反向传播求梯度
        opt.step()       #更新参数
        opt.zero_grad()  #梯度清零，避免影响下一次的更新参数
    return loss.item(), len(xb)

7.开始训练模型

train_dl,valid_dl=get_data(train_ds,valid_ds,bs)
model,opt=get_model()
fit(20,model ,loss_func,opt,train_dl,valid_dl)


correct=0
total=0
#xb(128,784) , yb(128)
for xb,yb in valid_dl:
    #output(128,10),每一批128个样例，10个概率
    output=(model(xb))
    # print(output.shape)
    # print(output)
    #predicted==预测概率中最大的值的索引
    _,predicted=torch.max(output.data,1) #最大的值和索引
    # print(predicted)
    #size(0)==64,item()脱离tensor
    total+=yb.size(0)
    correct+=(predicted==yb).sum().item()
print("Accuracy of network on the 10000 test image :%d %%" %(
    100*correct / total
))

plt.figure()
plt.title("loss during training")
plt.plot(np.arange(1,21,1),val_losses)
plt.show()

当前step:0 验证集损失0.19325110550522803
当前step:1 验证集损失0.13869898459613322
当前step:2 验证集损失0.11913147141262889
当前step:3 验证集损失0.10589157585203647
当前step:4 验证集损失0.09970801477096974
当前step:5 验证集损失0.09848284918610006
当前step:6 验证集损失0.08794679024070501
当前step:7 验证集损失0.08894123120522127
当前step:8 验证集损失0.0905570782547351
当前step:9 验证集损失0.0877237871955149
当前step:10 验证集损失0.08790379901565612
当前step:11 验证集损失0.08826288345884532
当前step:12 验证集损失0.08438722904250026
当前step:13 验证集损失0.08695273711904883
当前step:14 验证集损失0.08459821079988032
当前step:15 验证集损失0.08047270769253373
当前step:16 验证集损失0.0862937849830836
当前step:17 验证集损失0.08164657156261383
当前step:18 验证集损失0.08109720230847597
当前step:19 验证集损失0.08208743708985858
Accuracy of network on the 10000 test image :97 %