pytorch(一)：torch构建数据集并训练一个神经网络

目录

预备知识

导包

构建数据集

神经网络结构

训练测试精度可视化

计算模型精度

损失可视化

输出网络结构信息

训练神经网络

定义参数

载入数据

载入神经网络结构、损失及优化

训练及测试

损失、精度可视化

question

---

预备知识

1. pytorch(二):torch常用API应用及详解、torch.max()
2. 卷积层设置及输出大小计算
3. torch.utils.data.DataLoader使用方法
4. 加载手写数据集load_digits()
5. sklearn的train_test_split()的参数含义解释
6. torch.nn介绍
7. nn.Sequential
8. torch.nn.Conv2d()函数
9. relu()激活函数、relu图像、其他相关函数
10. nn.MaxPool2d()
11. nn.Linear()全连接层
12. x &＃61; x.view(x.size(0), -1)、Flatten
13. nn.CrossEntropyLoss()
14. torch.optim 优化器的使用
15. torch的Module模块中named_parameters函数
16. epoch
17. torch的net.train() 与net.eval()
18. enumerate(train_loader, start&＃61;0)函数的详解
19. python __getitem__()方法理解&＃xff08;扩充&＃xff09;
20. pytorch 损失函数及其应用代码详解、Pytorch十九种损失函数的使用、pytorch的nn.MSELoss损失函数
21. loss.backward() 
22. print()&＃43;end&＃61;""&＃43;\r实现滚动条显示
23. torch.no_grad和验证模式
24. (predicted &＃61;&＃61; labels).sum().item()作用
25. python基础之np.arange函数
26. Pytorch查看模型参数并计算模型参数量与可训练参数量

导包

import torch
from sklearn import datasets #加载数据集用
from sklearn import model_selection #数据集切分
import torch.utils.data as Data #分批次loader数据集使用
import torch.nn as nn #神经网络API
import torch.optim as optim #神经网络优化器
import numpy as np #数字格式转换
import matplotlib.pyplot as plt #数据可视化

构建数据集

class train_mini_train():&＃39;&＃39;&＃39;构建训练集&＃39;&＃39;&＃39;def __init__(self):# 加载手写数字数据集digits&＃xff0c;有标签&＃xff0c;class为1~10self.X,self.y &＃61; \datasets.load_digits(return_X_y&＃61;True)print(&＃39;样本总数&＃xff1a;&＃39;,len(self.X))print(&＃39;样本特征维度&＃xff1a;&＃39;, len(self.X[0]))#数据集切分self.X_train,self.X_test,self.y_train,self.y_test &＃61; \model_selection.train_test_split(self.X,self.y,random_state&＃61;0)print(&＃39;构建训练集样本总数&＃xff1a;&＃39;, len(self.y_train))def __len__(self):#返回训练集数据量return len(self.y_train)def __getitem__(self, index):return torch.tensor(self.X_train[index].reshape(1,8,8),dtype&＃61;torch.float32),self.y_train[index]class test_mini_test():&＃39;&＃39;&＃39;构建测试集&＃39;&＃39;&＃39;def __init__(self):self.X,self.y&＃61;datasets.load_digits(return_X_y&＃61;True)self.X_train,self.X_test,self.y_train,self.y_test &＃61;\model_selection.train_test_split(self.X,self.y,random_state&＃61;0)print(&＃39;构建测试集样本总数&＃xff1a;&＃39;, len(self.y_test))def __getitem__(self, index):return torch.tensor(self.X_test[index].reshape(1,8,8),dtype&＃61;torch.float32),self.y_test[index]def __len__(self):return len(self.y_test)

神经网络结构

class SimpleNet(nn.Module):&＃39;&＃39;&＃39;定义神经网络结构&＃39;&＃39;&＃39;def __init__(self):super(SimpleNet,self).__init__()self.conv1 &＃61; nn.Sequential(#(1, 8, 8)nn.Conv2d(in_channels&＃61;1,out_channels&＃61;4,kernel_size&＃61;2,stride&＃61;1,padding&＃61;1), #(4, 8, 8))self.conv2 &＃61; nn.Sequential(nn.ReLU(), # (4, 8, 8))self.conv3 &＃61; nn.Sequential(nn.MaxPool2d(kernel_size&＃61;2) # (4,4,4) 不改变通道数)self.conv4 &＃61; nn.Sequential(nn.Conv2d(in_channels&＃61;4,out_channels&＃61;8,kernel_size&＃61;2,stride&＃61;1,padding&＃61;1), #(8,5,5))self.conv5 &＃61; nn.Sequential(nn.ReLU(), # (8,5,5))self.conv6 &＃61; nn.Sequential(nn.MaxPool2d(kernel_size&＃61;2) # (8,2,2))self.fc &＃61; nn.Linear(8*2*2,10) #(10)def forward(self, x):x &＃61; self.conv1(x)x &＃61; self.conv2(x)x &＃61; self.conv3(x)x &＃61; self.conv4(x)x &＃61; self.conv5(x)x &＃61; self.conv6(x)x &＃61; x.view(x.size(0),-1) #相当于Flattenx &＃61; self.fc(x)return x

训练测试精度可视化

def plot_train_and_test_result(train_accs,test_accs):epoches &＃61; np.arange(1,len(train_accs)&＃43;1,dtype&＃61;np.int32)plt.plot(epoches,train_accs,label&＃61;&＃39;train accuracy&＃39;)plt.plot(epoches,test_accs, label&＃61;&＃39;test accuracy&＃39;)plt.xlabel(&＃39;epoches&＃39;)plt.ylabel(&＃39;accuracy&＃39;)plt.legend()

计算模型精度

def eval_on_dataloader(name,loader,len):acc &＃61; 0.0with torch.no_grad():for data in loader:images,labels &＃61; dataoutputs &＃61; net(images)# torch.max返回两个数值&＃xff0c;一个[0]是最大值&＃xff0c;一个[1]是最大值的下标predict_y &＃61; torch.max(outputs,dim&＃61;1)[1]acc &＃43;&＃61; (predict_y &＃61;&＃61; labels.long()).sum().item()accurate &＃61; acc/lenreturn accurate

损失可视化

def plot_loss_result(losses):epoches &＃61; np.arange(1, len(losses) &＃43; 1, dtype&＃61;np.int32)plt.plot(epoches, losses, label&＃61;&＃39;loss&＃39;)plt.xlabel(&＃39;epoches&＃39;)plt.ylabel(&＃39;loss&＃39;)plt.legend()

输出网络结构信息

def printNetInfo(net):&＃39;&＃39;&＃39;输出神经网络结构的信息&＃39;&＃39;&＃39;for name,parameters in net.named_parameters():print(name,":",parameters.size())

训练神经网络

定义参数

batch_size &＃61; 6
learning_rate &＃61; 3e-3
epoches &＃61; 10

载入数据

#载入训练集与测试集数据train_data &＃61; train_mini_train()test_data &＃61; test_mini_test()train_loader &＃61; Data.DataLoader(dataset&＃61;train_data,batch_size&＃61;batch_size,shuffle&＃61;True)test_loader &＃61; Data.DataLoader(dataset&＃61;test_data,batch_size&＃61;batch_size,shuffle&＃61;True)

载入神经网络结构、损失及优化

#载入网络结构net &＃61; SimpleNet()# 声明损失函数及优化器loss_fn &＃61; nn.CrossEntropyLoss()optim &＃61; optim.Adam(params&＃61;net.parameters(),lr&＃61;learning_rate)# 输出网络结构信息# printNetInfo(net)

训练及测试

train_accs &＃61; []test_accs &＃61; []losses &＃61; []for epoch in range(epoches):net.train() # 训练for step,data in enumerate(train_loader,start&＃61;0):images,labels &＃61; dataoptim.zero_grad() # 优化器梯度清0logits &＃61; net(images) # 输入images 经过网络推断输出结果loss &＃61; loss_fn(logits,labels.long()) # 计算损失函数loss.backward() # 反向传播求梯度optim.step() #优化器进一步优化rate &＃61; (step&＃43;1)/len(train_loader)a &＃61; "*"*int(rate*50)b &＃61; "."*int((1-rate)*50)print("\repoch:%s train loss:%3.0f%%:%.4f"%(epoch,int(rate*100),loss),end&＃61;" ")losses.append(loss)net.eval() # 测试train_acc &＃61; eval_on_dataloader("train",train_loader,train_data.__len__())train_accs.append(train_acc)test_acc &＃61; eval_on_dataloader("test", test_loader, test_data.__len__())test_accs.append(test_acc)print("train_acc:", train_acc, " test_acc:", test_acc)

损失、精度可视化

#模型训练损失可视化plot_loss_result(losses)plt.show()#训练集与测试集精度可视化plot_train_and_test_result(train_accs,test_accs)plt.show()

question

1. AttributeError: module ‘sklearn‘ has no attribute ‘datasets‘解决
2. RuntimeError: Expected object of scalar type Long but got scalar type Int for argument #2 &＃39;target&＃39;

solution&＃xff1a;data.long()数据类型转换