知识点回顾:
1.过拟合的判断:测试集和训练集同步打印指标
2.模型的保存和加载 a.仅保存权重 b.保存权重和模型 c.保存全部信息checkpoint,还包含训练状态
3.早停策略
作业:对信贷数据集训练后保存权重,加载权重后继续训练50轮,并采取早停策略
首先对信贷数据集进行训练后保存权重
import torch
import pandas as pd
import torch.nn as nn
import torch.optim as optim
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import MinMaxScaler
import time
import matplotlib.pyplot as plt
# 替换import numpy as np #用于数值计算,提供了高效的数组操作。
import matplotlib.pyplot as plt #用于绘制各种类型的图表
import seaborn as sns #基于matplotlib的高级绘图库,能绘制更美观的统计图形。
from tqdm import tqdm
# 设置GPU设备
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
print(f"使用设备: {device}")
# 设置中文字体(解决中文显示问题)
plt.rcParams['font.sans-serif'] = ['SimHei'] # Windows系统常用黑体字体
plt.rcParams['axes.unicode_minus'] = False # 正常显示负号
data = pd.read_csv(r'C:\Users\许兰\Desktop\打卡文件\python60-days-challenge-master\data.csv') #读取数据
# 先筛选字符串变量
discrete_features = data.select_dtypes(include=['object']).columns.tolist()
# Home Ownership 标签编码
home_ownership_mapping = {
'Own Home': 1,
'Rent': 2,
'Have Mortgage': 3,
'Home Mortgage': 4
}
data['Home Ownership'] = data['Home Ownership'].map(home_ownership_mapping)
# Years in current job 标签编码
years_in_job_mapping = {
'< 1 year': 1,
'1 year': 2,
'2 years': 3,
'3 years': 4,
'4 years': 5,
'5 years': 6,
'6 years': 7,
'7 years': 8,
'8 years': 9,
'9 years': 10,
'10+ years': 11
}
data['Years in current job'] = data['Years in current job'].map(years_in_job_mapping)
# Purpose 独热编码,记得需要将bool类型转换为数值
data = pd.get_dummies(data, columns=['Purpose'])
data2 = pd.read_csv(r'C:\Users\许兰\Desktop\打卡文件\python60-days-challenge-master\data.csv') # 重新读取数据,用来做列名对比
list_final = [] # 新建一个空列表,用于存放独热编码后新增的特征名
for i in data.columns:
if i not in data2.columns:
list_final.append(i) # 这里打印出来的就是独热编码后的特征名
for i in list_final:
data[i] = data[i].astype(int) # 这里的i就是独热编码后的特征名
# Term 0 - 1 映射
term_mapping = {
'Short Term': 0,
'Long Term': 1
}
data['Term'] = data['Term'].map(term_mapping)
data.rename(columns={'Term': 'Long Term'}, inplace=True) # 重命名列
continuous_features = data.select_dtypes(include=['int64', 'float64']).columns.tolist() #把筛选出来的列名转换成列表
# 连续特征用中位数补全
for feature in continuous_features:
mode_value = data[feature].mode()[0] #获取该列的众数。
data[feature].fillna(mode_value, inplace=True) #用众数填充该列的缺失值,inplace=True表示直接在原数据上
# 划分训练集、验证集和测试集,因为需要考2次
# 这里演示一下如何2次划分数据集,因为这个函数只能划分一次,所以需要调用两次才能划分出训练集、验证集和测试集。
from sklearn.model_selection import train_test_split
X = data.drop(['Credit Default'], axis=1) # 特征,axis=1表示按列删除
y = data['Credit Default'] # 标签
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42) # 80%训练集,20%临时集
# 归一化数据
scaler = MinMaxScaler()
X_train = scaler.fit_transform(X_train)
X_test = scaler.transform(X_test)
X_train = torch.FloatTensor(X_train).to(device)
y_train = torch.LongTensor(y_train.to_numpy()).to(device)
X_test = torch.FloatTensor(X_test).to(device)
y_test = torch.LongTensor(y_test).to(device)
class MLP(nn.Module):
def __init__(self):
super(MLP, self).__init__()
self.fc1 = nn.Linear(31, 18) # 输入层到隐藏层
self.relu = nn.ReLU()
self.fc2 = nn.Linear(18, 2) # 隐藏层到输出层
def forward(self, x):
out = self.fc1(x)
out = self.relu(out)
out = self.fc2(out)
return out
# 实例化模型并移至GPU
model = MLP()
# 分类问题使用交叉熵损失函数
criterion = nn.CrossEntropyLoss()
# 使用随机梯度下降优化器
optimizer = optim.SGD(model.parameters(), lr=0.01)
# 训练模型
num_epochs = 20000 # 训练的轮数
# 用于存储每100个epoch的损失值和对应的epoch数
train_losses = []
test_losses = []
epochs = []
# ===== 新增早停相关参数 =====
best_test_loss = float('inf') # 记录最佳测试集损失
best_epoch = 0 # 记录最佳epoch
patience = 50 # 早停耐心值(连续多少轮测试集损失未改善时停止训练)
counter = 0 # 早停计数器
early_stopped = False
start_time = time.time() # 记录开始时间
# 创建tqdm进度条
with tqdm(total=num_epochs, desc="训练进度", unit="epoch") as pbar:
# 训练模型
for epoch in range(num_epochs):
# 前向传播
outputs = model(X_train) # 隐式调用forward函数
train_loss = criterion(outputs, y_train)
# 反向传播和优化
optimizer.zero_grad()
train_loss.backward()
optimizer.step()
# 记录损失值并更新进度条
if (epoch + 1) % 200 == 0:
model.eval()
with torch.no_grad():
test_outputs = model(X_test)
test_loss = criterion(test_outputs, y_test)
model.train()
train_losses.append(train_loss.item())
test_losses.append(test_loss.item())
epochs.append(epoch + 1)
# 更新进度条的描述信息
pbar.set_postfix({'Train Loss': f'{train_loss.item():.4f}', 'Test Loss': f'{test_loss.item():.4f}'})
# ===== 新增早停逻辑 =====
if test_loss.item() < best_test_loss: # 如果当前测试集损失小于最佳损失
best_test_loss = test_loss.item() # 更新最佳损失
best_epoch = epoch + 1 # 更新最佳epoch
counter = 0 # 重置计数器
# 保存最佳模型
torch.save(model.state_dict(), 'best_model.pth')
else:
counter += 1
if counter >= patience:
print(f"早停触发!在第{epoch+1}轮,测试集损失已有{patience}轮未改善。")
print(f"最佳测试集损失出现在第{best_epoch}轮,损失值为{best_test_loss:.4f}")
early_stopped = True
break # 终止训练循环
# ======================
# 每1000个epoch更新一次进度条
if (epoch + 1) % 1000 == 0:
pbar.update(1000) # 更新进度条
# 确保进度条达到100%
if pbar.n < num_epochs:
pbar.update(num_epochs - pbar.n) # 计算剩余的进度并更新
time_all = time.time() - start_time # 计算训练时间
print(f'Training time: {time_all:.2f} seconds')
# ===== 新增:加载最佳模型用于最终评估 =====
if early_stopped:
print(f"加载第{best_epoch}轮的最佳模型进行最终评估...")
model.load_state_dict(torch.load('best_model.pth'))
# 可视化损失曲线
plt.figure(figsize=(10, 6))
plt.plot(epochs, train_losses, label='Train Loss') # 原始代码已有
plt.plot(epochs, test_losses, label='Test Loss') # 新增:测试集损失曲线
plt.xlabel('Epoch')
plt.ylabel('Loss')
plt.title('Training and Test Loss over Epochs')
plt.legend() # 新增:显示图例
plt.grid(True)
plt.show()
model.eval() # 设置模型为评估模式
with torch.no_grad(): # torch.no_grad()的作用是禁用梯度计算,可以提高模型推理速度
outputs = model(X_test) # 对测试数据进行前向传播,获得预测结果
_, predicted = torch.max(outputs, 1) # torch.max(outputs, 1)返回每行的最大值和对应的索引
correct = (predicted == y_test).sum().item() # 计算预测正确的样本数
accuracy = correct / y_test.size(0)
print(f'测试集准确率: {accuracy * 100:.2f}%')
torch.save(model.state_dict(), "model_weights.pth")
保存权重后的模型损失曲线
测试集准确率
采取早停策略后
并没有触发早停策略