4.2 面向对象编程 - AI应用开发

🎯 学习目标

理解类与对象的概念
掌握继承与多态
学会设计可复用的类
了解PyTorch中的OOP模式

图：面向对象是组织复杂代码的有效方式

🏗️ 类的定义

class NeuralNetwork:
    """神经网络基类"""

    # 类属性
    version = "1.0"

    def __init__(self, input_size, hidden_size, output_size):
        # 实例属性
        self.input_size = input_size
        self.hidden_size = hidden_size
        self.output_size = output_size
        self.weights = None

    def forward(self, x):
        """前向传播"""
        raise NotImplementedError

    def __repr__(self):
        return f"NeuralNetwork(input={self.input_size}, hidden={self.hidden_size})"

    def __len__(self):
        return self.input_size + self.hidden_size + self.output_size

# 使用
model = NeuralNetwork(784, 256, 10)
print(model)  # NeuralNetwork(input=784, hidden=256)
print(len(model))  # 1030
      

🔄 继承与多态

继承示例

import torch
import torch.nn as nn

# 继承PyTorch的Module类
class MLP(nn.Module):
    def __init__(self, input_size, hidden_size, output_size):
        super().__init__()  # 调用父类初始化

        self.fc1 = nn.Linear(input_size, hidden_size)
        self.fc2 = nn.Linear(hidden_size, output_size)
        self.relu = nn.ReLU()

    def forward(self, x):
        x = self.fc1(x)
        x = self.relu(x)
        x = self.fc2(x)
        return x

# 多继承
class Mixin:
    def save(self, path):
        torch.save(self.state_dict(), path)

class MyModel(MLP, Mixin):
    pass

model = MyModel(784, 256, 10)
model.save("model.pt")
        

🎭 属性装饰器

class Dataset:
    def __init__(self, data):
        self._data = data
        self._size = None

    @property
    def size(self):
        """延迟计算大小"""
        if self._size is None:
            self._size = len(self._data)
        return self._size

    @size.setter
    def size(self, value):
        raise AttributeError("Cannot set size directly")

    @staticmethod
    def from_file(path):
        """静态方法：从文件创建数据集"""
        with open(path) as f:
            data = f.readlines()
        return Dataset(data)

    @classmethod
    def empty(cls):
        """类方法：创建空数据集"""
        return cls([])

# 使用
ds = Dataset([1, 2, 3, 4, 5])
print(ds.size)  # 5

ds2 = Dataset.from_file("data.txt")
ds3 = Dataset.empty()
      

📦 数据类（Dataclass）

Python 3.7+ 简化类定义

from dataclasses import dataclass, field
from typing import List

@dataclass
class TrainingConfig:
    learning_rate: float = 0.001
    batch_size: int = 32
    epochs: int = 100
    optimizer: str = "adam"
    layers: List[int] = field(default_factory=lambda: [256, 128])

    def __post_init__(self):
        # 初始化后执行
        if self.learning_rate <= 0:
            raise ValueError("learning_rate must be positive")

# 使用
config = TrainingConfig(learning_rate=0.0001, epochs=50)
print(config)
# TrainingConfig(learning_rate=0.0001, batch_size=32,
#                epochs=50, optimizer='adam', layers=[256, 128])
        

📝 本节小结

✅

• 类是创建对象的模板，封装数据和行为
• 继承实现代码复用，多态实现接口统一
• @property实现属性的getter/setter
• @dataclass简化数据类的定义