real_training_wrapper.py

#!/usr/bin/env python3
# -*- coding: utf-8 -*-

"""
真实训练包装器代码
将此代码集成到 task_evaluator.py 中
"""

from typing import Dict, Any, Tuple
import argparse


def execute_real_training_with_progress(self, parameters: Dict[str, Any], args: argparse.Namespace,
                                      fold: int, total_epochs: int) -> Tuple[float, float, float, float, float, float]:
    """
    执行真实训练的包装器函数
    """
    from nested_progress_manager import update_epoch_progress
    
    try:
        # 尝试执行真实训练
        return self._execute_real_training(parameters, args, fold, total_epochs)
    except Exception as e:
        self.logger.error(f"真实训练失败，回退到模拟训练: {str(e)}")
        return self._fallback_simulate_training(parameters, args, fold, total_epochs)

def _execute_real_training(self, parameters: Dict[str, Any], args: argparse.Namespace,
                          fold: int, total_epochs: int) -> Tuple[float, float, float, float, float, float]:
    """
    执行真实的神经网络训练
    """
    # 修复参数问题
    from fix_training_issues import TrainingIssueFixer
    fixer = TrainingIssueFixer()
    fixed_params, was_fixed = fixer.validate_and_fix_parameters(parameters)
    
    if was_fixed:
        # 更新args中的参数
        for key, value in fixed_params.items():
            setattr(args, key, value)
    
    # 导入训练模块
    from train import train_model
    import torch
    import torch.optim as optim
    
    # 设置设备
    device = 'cuda' if torch.cuda.is_available() else 'cpu'
    
    # 创建模型（这里需要根据实际情况调整）
    try:
        from layer import GCN  # 假设这是您的模型类
        model = GCN(args).to(device)
    except Exception as e:
        self.logger.error(f"模型创建失败: {e}")
        raise
    
    # 创建优化器
    optimizer = optim.Adam(model.parameters(), lr=args.lr, weight_decay=args.weight_decay)
    
    # 准备数据
    data_o, data_a, train_loader, test_loader = self._prepare_training_data(args, fold)
    
    # 执行真实训练
    results = train_model(model, optimizer, data_o, data_a, train_loader, test_loader, args, fold)
    
    return (
        results['auroc'],
        results['auprc'], 
        results['f1'],
        results['precision'],
        results['recall'],
        results['loss']
    )

def _prepare_training_data(self, args: argparse.Namespace, fold: int):
    """
    准备训练数据
    """
    # 这里需要根据您的实际数据加载逻辑来实现
    # 暂时抛出异常，提醒需要实现
    raise NotImplementedError("需要实现数据加载逻辑")