test/test_complete_pipeline.py

#!/usr/bin/env python3
"""
Complete TestTime RLVR Pipeline Test Script

AZR 기반 TestTime RLVR 파이프라인을 실제 벤치마크 문제로 테스트
LLM 솔루션 생성 → IPO 추출 → 태스크 생성 → LLM 평가 → Reward 계산 전체 플로우 검증
"""

import os
import sys
import torch
import argparse
import json
from pathlib import Path
from datetime import datetime

# TestTime RLVR 모듈 임포트
sys.path.append('/home/ubuntu/RLVR/TestTime-RLVR-v2')
from absolute_zero_reasoner.testtime.complete_pipeline import CompleteTestTimePipeline
from absolute_zero_reasoner.testtime.config import TestTimeConfig, BenchmarkConfig
from absolute_zero_reasoner.testtime.logger import TestTimeLogger
from absolute_zero_reasoner.testtime.solution_generator import InitialSolutionGenerator


def load_test_problem():
    """간단한 테스트 문제 생성 (HumanEval 스타일)"""
    return {
        'task_id': 'test/simple_sum',
        'prompt': '''def add_two_numbers(a, b):
    """
    Add two numbers and return the result.
    
    Args:
        a (int): First number
        b (int): Second number
    
    Returns:
        int: Sum of a and b
    
    Examples:
        >>> add_two_numbers(2, 3)
        5
        >>> add_two_numbers(-1, 1)
        0
        >>> add_two_numbers(0, 0)
        0
    """''',
        'entry_point': 'add_two_numbers',
        'canonical_solution': 'def add_two_numbers(a, b):\n    return a + b',
        'test': '''def check(candidate):
    assert candidate(2, 3) == 5
    assert candidate(-1, 1) == 0
    assert candidate(0, 0) == 0
    assert candidate(10, -5) == 5'''
    }


def save_detailed_results(result, args, output_dir):
    """상세한 결과를 개별 파일로 저장"""
    
    # 벤치마크와 문제 ID에 따른 디렉토리 구조 생성
    benchmark = result.get('benchmark', 'unknown')
    problem_id = result['problem_id']  # '/' 유지
    problem_id_safe = problem_id.replace('/', '_')  # 파일명용
    
    # {output_dir}/{benchmark}/{task_id} 구조로 디렉토리 생성
    base_dir = os.path.join(output_dir, benchmark, problem_id_safe)
    os.makedirs(base_dir, exist_ok=True)
    
    timestamp = datetime.now().strftime('%Y%m%d_%H%M%S')
    
    # 1. 초기 LLM 솔루션 저장 (벤치마크 문제 해결)
    if 'llm_generation' in result['steps']:
        llm_step = result['steps']['llm_generation']
        
        initial_solution_dir = os.path.join(base_dir, 'initial_solution')
        os.makedirs(initial_solution_dir, exist_ok=True)
        
        # 벤치마크 문제 원본 저장
        if 'problem_loading' in result['steps']:
            problem_data = result['steps']['problem_loading'].get('problem', {})
            problem_file = os.path.join(initial_solution_dir, f"{problem_id_safe}_original_problem.txt")
            with open(problem_file, 'w', encoding='utf-8') as f:
                f.write(f"Problem ID: {result['problem_id']}\n")
                f.write(f"Benchmark: {result['benchmark']}\n")
                f.write(f"Generated: {timestamp}\n")
                f.write("="*80 + "\n")
                f.write("ORIGINAL BENCHMARK PROBLEM:\n")
                f.write("="*80 + "\n")
                f.write(problem_data.get('prompt', 'No prompt available'))
                f.write("\n" + "="*80 + "\n")
                f.write("ENTRY POINT:\n")
                f.write("="*80 + "\n")
                f.write(problem_data.get('entry_point', 'No entry point'))
                f.write("\n" + "="*80 + "\n")
                f.write("CANONICAL SOLUTION:\n")
                f.write("="*80 + "\n")
                f.write(problem_data.get('canonical_solution', 'No canonical solution'))
                if 'test' in problem_data:
                    f.write("\n" + "="*80 + "\n")
                    f.write("TEST CASES:\n")
                    f.write("="*80 + "\n")
                    f.write(str(problem_data['test']))
        
        # LLM 생성 솔루션 저장
        llm_solution_file = os.path.join(initial_solution_dir, f"{problem_id_safe}_llm_solution.txt")
        with open(llm_solution_file, 'w', encoding='utf-8') as f:
            f.write(f"Problem ID: {result['problem_id']}\n")
            f.write(f"Benchmark: {result['benchmark']}\n")
            f.write(f"Generated: {timestamp}\n")
            f.write("="*80 + "\n")
            f.write("LLM GENERATED SOLUTION:\n")
            f.write("="*80 + "\n")
            f.write(llm_step.get('solution', 'No solution generated'))
            f.write("\n" + "="*80 + "\n")
            f.write("SYNTAX VALIDATION:\n")
            f.write("="*80 + "\n")
            syntax_valid = llm_step.get('syntax_valid', False)
            f.write(f"Valid: {'✅ YES' if syntax_valid else '❌ NO'}")
            if llm_step.get('syntax_error'):
                f.write(f"\nError: {llm_step['syntax_error']}")
            
            # 초기 솔루션 정확성 평가 결과 추가
            f.write("\n" + "="*80 + "\n")
            f.write("SOLUTION CORRECTNESS EVALUATION:\n")
            f.write("="*80 + "\n")
            
            solution_eval = llm_step.get('solution_evaluation')
            if solution_eval:
                if solution_eval['correct']:
                    f.write(f"Result: ✅ CORRECT ({solution_eval['passed_tests']}/{solution_eval['total_tests']} tests passed)\n")
                else:
                    f.write(f"Result: ❌ INCORRECT ({solution_eval['passed_tests']}/{solution_eval['total_tests']} tests passed)\n")
                
                if solution_eval.get('error'):
                    f.write(f"Error: {solution_eval['error']}\n")
                
                # 실행 결과 상세 정보
                if solution_eval.get('execution_results'):
                    f.write("\nExecution Details:\n")
                    for i, exec_result in enumerate(solution_eval['execution_results']):
                        f.write(f"  Test {i+1}:\n")
                        f.write(f"    Status: {exec_result.get('status', 'N/A')}\n")
                        if 'result' in exec_result:
                            f.write(f"    Result: {exec_result['result'][:100]}...\n")
            else:
                f.write("No evaluation performed (syntax error or no test cases)\n")
        
        # IPO 추출을 위해 사용된 프로그램 저장
        if 'ipo_extraction' in result['steps']:
            ipo_step = result['steps']['ipo_extraction']
            if 'extracted_program' in ipo_step:
                extracted_program_file = os.path.join(initial_solution_dir, f"{problem_id_safe}_extracted_program.py")
                with open(extracted_program_file, 'w', encoding='utf-8') as f:
                    f.write(f"# Problem ID: {result['problem_id']}\n")
                    f.write(f"# Benchmark: {result['benchmark']}\n")
                    f.write(f"# Generated: {timestamp}\n")
                    f.write(f"# Extracted from LLM solution for IPO generation\n\n")
                    f.write(ipo_step['extracted_program'])
                    
        print(f"📁 초기 솔루션 저장: {initial_solution_dir}/")
    
    # 2. IPO 트리플 저장
    if 'ipo_extraction' in result['steps']:
        ipo_step = result['steps']['ipo_extraction']
        triples = ipo_step.get('triples', [])
        
        ipo_dir = os.path.join(base_dir, 'ipo_triples')
        os.makedirs(ipo_dir, exist_ok=True)
        
        for i, triple in enumerate(triples):
            triple_file = os.path.join(ipo_dir, f"{problem_id_safe}_triple_{i+1}.json")
            with open(triple_file, 'w', encoding='utf-8') as f:
                json.dump(triple, f, indent=2, ensure_ascii=False)
                
        print(f"📁 IPO 트리플 저장: {ipo_dir}/ ({len(triples)}개 파일)")
    
    # 3. 생성된 태스크 프롬프트 저장
    if 'task_generation' in result['steps']:
        task_step = result['steps']['task_generation']
        all_tasks = task_step.get('all_tasks', {})
        
        task_dir = os.path.join(base_dir, 'task_prompts')
        os.makedirs(task_dir, exist_ok=True)
        
        task_count = 0
        for task_type, tasks in all_tasks.items():
            for i, task in enumerate(tasks):
                task_file = os.path.join(task_dir, f"{problem_id_safe}_{task_type}_{i+1}.txt")
                with open(task_file, 'w', encoding='utf-8') as f:
                    f.write(f"Task Type: {task_type}\n")
                    f.write(f"Task ID: {task.get('task_id', 'N/A')}\n")
                    f.write(f"Generated: {timestamp}\n")
                    f.write("="*80 + "\n")
                    f.write("TASK PROMPT:\n")
                    f.write("="*80 + "\n")
                    f.write(task.get('prompt', 'No prompt available'))
                    f.write("\n" + "="*80 + "\n")
                    f.write("EXPECTED SOLUTION:\n")
                    f.write("="*80 + "\n")
                    f.write(task.get('expected_solution', 'No expected solution'))
                    f.write("\n" + "="*80 + "\n")
                    f.write("EVALUATION DATA:\n")
                    f.write("="*80 + "\n")
                    f.write(str(task.get('evaluation_data', 'No evaluation data')))
                task_count += 1
                
        print(f"📁 태스크 프롬프트 저장: {task_dir}/ ({task_count}개 파일)")
    
    # 4. LLM 태스크 응답 저장
    if 'task_evaluation' in result['steps']:
        eval_step = result['steps']['task_evaluation']
        evaluations = eval_step.get('evaluations', {})
        
        response_dir = os.path.join(base_dir, 'llm_responses')
        os.makedirs(response_dir, exist_ok=True)
        
        response_count = 0
        for task_type, task_evals in evaluations.items():
            for i, evaluation in enumerate(task_evals):
                response_file = os.path.join(response_dir, f"{problem_id_safe}_{task_type}_{i+1}_response.txt")
                with open(response_file, 'w', encoding='utf-8') as f:
                    f.write(f"Task Type: {task_type}\n")
                    f.write(f"Task ID: {evaluation.get('task_id', 'N/A')}\n")
                    f.write(f"Generated: {timestamp}\n")
                    f.write("="*80 + "\n")
                    f.write("ORIGINAL PROMPT:\n")
                    f.write("="*80 + "\n")
                    f.write(evaluation.get('prompt', 'No prompt available'))
                    f.write("\n" + "="*80 + "\n")
                    f.write("LLM RESPONSE:\n")
                    f.write("="*80 + "\n")
                    f.write(evaluation.get('llm_response', 'No response'))
                    f.write("\n" + "="*80 + "\n")
                    f.write("EXPECTED SOLUTION:\n")
                    f.write("="*80 + "\n")
                    f.write(evaluation.get('expected_solution', 'No expected solution'))
                    
                    # 추출된 정답 정보 추가 (보상 계산 결과에서 가져오기)
                    if 'reward_computation' in result['steps']:
                        reward_step = result['steps']['reward_computation']
                        rewards = reward_step.get('rewards', {})
                        rewards_by_type = rewards.get('rewards_by_type', {})
                        
                        # 현재 태스크의 보상 정보 찾기
                        current_task_rewards = rewards_by_type.get(task_type, [])
                        current_reward = None
                        for reward in current_task_rewards:
                            if reward.get('task_id') == evaluation.get('task_id'):
                                current_reward = reward
                                break
                        
                        if current_reward and 'extracted_answer' in current_reward:
                            f.write("\n" + "="*80 + "\n")
                            f.write("EXTRACTED ANSWER:\n")
                            f.write("="*80 + "\n")
                            f.write(current_reward['extracted_answer'])
                            f.write("\n" + "="*80 + "\n")
                            f.write("MATCH RESULT:\n")
                            f.write("="*80 + "\n")
                            match_result = "✅ CORRECT" if current_reward.get('basic_accuracy', 0) > 0 else "❌ INCORRECT"
                            f.write(f"{match_result} (Score: {current_reward.get('basic_accuracy', 0):.3f})")
                            
                response_count += 1
                
        print(f"📁 LLM 응답 저장: {response_dir}/ ({response_count}개 파일)")
    
    # 4-1. 추출된 정답 별도 저장
    if 'reward_computation' in result['steps']:
        reward_step = result['steps']['reward_computation']
        rewards = reward_step.get('rewards', {})
        rewards_by_type = rewards.get('rewards_by_type', {})
        
        extracted_dir = os.path.join(base_dir, 'extracted_answers')
        os.makedirs(extracted_dir, exist_ok=True)
        
        extracted_count = 0
        for task_type, task_rewards in rewards_by_type.items():
            for reward in task_rewards:
                if 'extracted_answer' in reward:
                    task_id = reward.get('task_id', 'unknown')
                    extracted_file = os.path.join(extracted_dir, f"{problem_id_safe}_{task_type}_{task_id}_extracted.txt")
                    with open(extracted_file, 'w', encoding='utf-8') as f:
                        f.write(f"Task Type: {task_type}\n")
                        f.write(f"Task ID: {task_id}\n")
                        f.write(f"Generated: {timestamp}\n")
                        f.write("="*80 + "\n")
                        f.write("EXTRACTED ANSWER:\n")
                        f.write("="*80 + "\n")
                        f.write(reward['extracted_answer'])
                        f.write("\n" + "="*80 + "\n")
                        f.write("EXPECTED SOLUTION:\n")
                        f.write("="*80 + "\n")
                        f.write(reward['expected_solution'])
                        f.write("\n" + "="*80 + "\n")
                        f.write("MATCH RESULT:\n")
                        f.write("="*80 + "\n")
                        match_result = "✅ CORRECT" if reward.get('basic_accuracy', 0) > 0 else "❌ INCORRECT"
                        f.write(f"{match_result} (Score: {reward.get('basic_accuracy', 0):.3f})")
                    extracted_count += 1
                    
        print(f"📁 추출된 정답 저장: {extracted_dir}/ ({extracted_count}개 파일)")
    
    # 5. 정답 비교 및 보상 결과 저장
    if 'reward_computation' in result['steps']:
        reward_step = result['steps']['reward_computation']
        rewards = reward_step.get('rewards', {})
        
        reward_file = os.path.join(base_dir, f"{problem_id_safe}_reward_analysis.json")
        with open(reward_file, 'w', encoding='utf-8') as f:
            json.dump(rewards, f, indent=2, ensure_ascii=False)
            
        # 사람이 읽기 쉬운 보상 요약 저장
        summary_file = os.path.join(base_dir, f"{problem_id_safe}_reward_summary.txt")
        with open(summary_file, 'w', encoding='utf-8') as f:
            f.write(f"REWARD ANALYSIS SUMMARY\n")
            f.write(f"Problem: {result['problem_id']}\n")
            f.write(f"Benchmark: {result['benchmark']}\n")
            f.write(f"Generated: {timestamp}\n")
            f.write("="*80 + "\n")
            
            f.write(f"OVERALL STATISTICS:\n")
            f.write(f"- Total Tasks: {rewards.get('total_tasks', 0)}\n")
            f.write(f"- Average Reward: {rewards.get('average_reward', 0.0):.3f}\n")
            f.write("\n")
            
            f.write(f"REWARD BY TASK TYPE:\n")
            for task_type, avg_reward in rewards.get('reward_distribution', {}).items():
                f.write(f"- {task_type.title()}: {avg_reward:.3f}\n")
            f.write("\n")
            
            f.write(f"DETAILED TASK REWARDS:\n")
            for task_type, task_rewards in rewards.get('rewards_by_type', {}).items():
                f.write(f"\n{task_type.upper()} TASKS:\n")
                for reward in task_rewards:
                    f.write(f"  Task {reward['task_id']}: ")
                    f.write(f"Accuracy={reward['basic_accuracy']:.3f}, ")
                    f.write(f"Final={reward['final_reward']:.3f}\n")
                    
        print(f"📁 보상 분석 저장: {reward_file}")
        print(f"📁 보상 요약 저장: {summary_file}")
    
    # 6. 전체 결과 요약 저장 (JSON 직렬화 가능하게 수정)
    summary_file = os.path.join(base_dir, f"{problem_id_safe}_pipeline_summary.json")
    
    # JSON 직렬화 가능하도록 결과 정리
    serializable_result = result.copy()
    
    # BenchmarkConfig 객체 제거 또는 직렬화 가능한 형태로 변환
    if 'steps' in serializable_result and 'problem_loading' in serializable_result['steps']:
        problem_data = serializable_result['steps']['problem_loading'].get('problem', {})
        if 'benchmark_config' in problem_data:
            # BenchmarkConfig 객체를 딕셔너리로 변환
            config_obj = problem_data['benchmark_config']
            problem_data['benchmark_config'] = {
                'name': config_obj.name,
                'data_path': config_obj.data_path,
                'problem_prefix': config_obj.problem_prefix,
                'max_problems': config_obj.max_problems,
                'test_timeout': config_obj.test_timeout
            }
    
    with open(summary_file, 'w', encoding='utf-8') as f:
        json.dump(serializable_result, f, indent=2, ensure_ascii=False)
        
    print(f"📁 전체 결과 요약 저장: {summary_file}")
    print(f"\n📂 모든 결과 파일 저장 완료: {output_dir}")


def main():
    parser = argparse.ArgumentParser(description='Test Complete TestTime RLVR Pipeline')
    parser.add_argument('--model', type=str, default='Qwen/Qwen2.5-7B', 
                       help='Model name to test with')
    parser.add_argument('--gpu', type=int, default=0, help='GPU ID to use')
    parser.add_argument('--max_tokens', type=int, default=512, help='Max tokens for generation')
    parser.add_argument('--benchmark', type=str, default='test', 
                       choices=['test', 'humaneval', 'mbpp'],
                       help='Benchmark to use (test=example data, humaneval=HumanEval+, mbpp=MBPP+)')
    parser.add_argument('--problem_id', type=str, default='test/simple_sum',
                       help='Problem ID to test (e.g., HumanEval/0, Mbpp/2)')
    parser.add_argument('--output_dir', type=str, default='../tmp',
                       help='Output directory for detailed results')
    parser.add_argument('--verbose', action='store_true', help='Verbose logging')
    
    args = parser.parse_args()
    
    # GPU 설정
    device = f'cuda:{args.gpu}' if torch.cuda.is_available() else 'cpu'
    print(f"🎯 Using device: {device}")
    
    # TestTime 설정
    config = TestTimeConfig(
        model_name=args.model,
        max_adaptation_steps=3,
        learning_rate=1e-5,
        task_distribution={'induction': 0.4, 'deduction': 0.3, 'abduction': 0.3},
        adaptation_batch_size=1,
        max_tasks_per_type=3,
        use_flash_attention=False,  # 작은 모델에서는 비활성화
        torch_dtype=torch.float16,
        enable_gradient_checkpointing=False
    )
    
    # 벤치마크 설정 (절대 경로로 계산)
    base_dir = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
    
    if args.benchmark == 'humaneval':
        benchmark_config = BenchmarkConfig.get_humaneval_config()
        benchmark_config.data_path = os.path.join(base_dir, 'evaluation/code_eval/data/HumanEvalPlus.jsonl')
    elif args.benchmark == 'mbpp':
        benchmark_config = BenchmarkConfig.get_mbpp_config()
        benchmark_config.data_path = os.path.join(base_dir, 'evaluation/code_eval/data/MbppPlus.jsonl')
    else:  # test
        benchmark_config = BenchmarkConfig(
            name='test_humaneval',
            data_path='test_data',
            problem_prefix='TestEval',
            max_problems=1,
            test_timeout=30
        )
    
    # 로거 설정
    logger = TestTimeLogger(log_level='DEBUG' if args.verbose else 'INFO')
    logger.log_info("🚀 Starting Complete TestTime RLVR Pipeline Test")
    logger.log_info(f"📋 Model: {args.model}")
    logger.log_info(f"🎯 Device: {device}")
    
    try:
        # 모델 및 토크나이저 로드 (VLLM 최적화 적용)
        logger.log_info("📦 Loading model and tokenizer with VLLM optimization...")
        model, tokenizer = InitialSolutionGenerator.load_model_with_optimizations(
            args.model, device, config, use_vllm=True  # VLLM 최적화 활성화
        )
        logger.log_info("✅ Model loaded successfully")
        
        # 파이프라인 초기화
        logger.log_info("🔧 Initializing pipeline...")
        pipeline = CompleteTestTimePipeline(model, tokenizer, config, logger)
        logger.log_info("✅ Pipeline initialized")
        
        # 문제 ID 설정
        problem_id = args.problem_id
        
        logger.log_info(f"📄 Testing with {args.benchmark} benchmark")
        logger.log_info(f"🔍 Problem ID: {problem_id}")
        
        # 테스트 모드인 경우 예시 데이터 사용
        if args.benchmark == 'test':
            test_problem = load_test_problem()
            logger.log_info(f"🔍 Problem preview: {test_problem['prompt'][:100]}...")
            # 임시로 문제를 pipeline의 benchmark_loader에 직접 제공
            pipeline.benchmark_loader.load_problem = lambda cfg, pid: test_problem
        else:
            # 실제 벤치마크 사용 시 프롬프트 미리보기
            temp_problem = pipeline.benchmark_loader.load_problem(benchmark_config, problem_id)
            
            # AZR 코드 평가 프롬프트 포맷 적용
            azr_prompt = f"Please provide a self-contained Python script that solves the following problem in a markdown code block:\n\n{temp_problem.get('prompt', 'No prompt available')}"
            
            print(f"\n📋 **ORIGINAL PROBLEM:**")
            print("="*80)
            print(temp_problem.get('prompt', 'No prompt available'))
            print("="*80)
            
            print(f"\n📋 **AZR CODE EVALUATION PROMPT (실제 사용되는 프롬프트):**")
            print("="*80)
            print(azr_prompt)
            print("="*80)
            
            print(f"📌 Entry Point: {temp_problem.get('entry_point', 'N/A')}")
            print(f"📌 Task ID: {temp_problem.get('task_id', 'N/A')}")
            if 'test' in temp_problem:
                print(f"📌 Test Preview: {str(temp_problem['test'])[:200]}...")
            print("="*80)
        
        # 전체 파이프라인 실행
        logger.log_info("🏃‍♂️ Running complete pipeline...")
        print("\n" + "="*60)
        print("🚀 COMPLETE TESTTIME RLVR PIPELINE EXECUTION")
        print(f"📋 Benchmark: {args.benchmark}")
        print(f"🔍 Problem: {problem_id}")
        print("="*60)
        
        result = pipeline.run_complete_pipeline(benchmark_config, problem_id)
        
        print("\n" + "="*60)
        print("📊 PIPELINE EXECUTION RESULTS")
        print("="*60)
        
        # 결과 출력
        print(f"✅ Success: {result['success']}")
        if result['error']:
            print(f"❌ Error: {result['error']}")
        
        print(f"📋 Problem: {result['problem_id']}")
        print(f"🏷️  Benchmark: {result['benchmark']}")
        
        # 단계별 결과 출력
        for step_name, step_result in result['steps'].items():
            print(f"\n📍 Step: {step_name.replace('_', ' ').title()}")
            print(f"   Success: {'✅' if step_result['success'] else '❌'}")
            
            if step_name == 'llm_generation':
                solution = step_result.get('solution', '')
                print(f"   Solution preview: {solution[:100]}...")
                print(f"   Syntax valid: {'✅' if step_result.get('syntax_valid') else '❌'}")
                
                # 초기 솔루션 정확성 평가 결과 표시
                eval_result = step_result.get('solution_evaluation')
                if eval_result:
                    if eval_result['correct']:
                        print(f"   ✅ Solution CORRECT ({eval_result['passed_tests']}/{eval_result['total_tests']} tests passed)")
                    else:
                        print(f"   ❌ Solution INCORRECT ({eval_result['passed_tests']}/{eval_result['total_tests']} tests passed)")
                        if eval_result.get('error'):
                            print(f"      Error: {eval_result['error'][:80]}...")
                
            elif step_name == 'ipo_extraction':
                print(f"   IPO triples extracted: {step_result.get('num_triples', 0)}")
                
            elif step_name == 'task_generation':
                print(f"   Total tasks generated: {step_result.get('total_tasks', 0)}")
                for task_type, count in step_result.get('tasks_by_type', {}).items():
                    print(f"     {task_type}: {count}")
                    
            elif step_name == 'task_evaluation':
                evaluations = step_result.get('evaluations', {})
                total_evaluated = sum(len(evals) for evals in evaluations.values())
                print(f"   Tasks evaluated: {total_evaluated}")
                
            elif step_name == 'reward_computation':
                rewards = step_result.get('rewards', {})
                print(f"   Average reward: {rewards.get('average_reward', 0.0):.3f}")
                print(f"   Total tasks scored: {rewards.get('total_tasks', 0)}")
                
                # 정답 추출 상세 정보 표시
                for task_type, type_rewards in rewards.get('rewards_by_type', {}).items():
                    print(f"   {task_type.title()} Tasks:")
                    for reward in type_rewards[:2]:  # 처음 2개만 표시
                        print(f"     Task {reward['task_id']}: Expected='{reward['expected_solution'][:50]}...' | Extracted='{reward['extracted_answer'][:50]}...' | Match={'✅' if reward['basic_accuracy'] > 0 else '❌'}")
        
        # 상세 결과 표시 (verbose 모드)
        if args.verbose and result['success']:
            print("\n" + "="*60)
            print("🔍 DETAILED RESULTS (VERBOSE MODE)")
            print("="*60)
            
            # IPO 추출 상세
            if 'ipo_extraction' in result['steps']:
                ipo_step = result['steps']['ipo_extraction']
                triples = ipo_step.get('triples', [])
                print(f"\n📊 IPO Triples ({len(triples)}):")
                for i, triple in enumerate(triples[:3]):  # 처음 3개만 표시
                    print(f"   [{i+1}] Input: {str(triple.get('input', 'N/A'))[:50]}...")
                    print(f"       Output: {str(triple.get('output', 'N/A'))[:50]}...")
            
            # 태스크 생성 상세
            if 'task_generation' in result['steps']:
                task_step = result['steps']['task_generation']
                all_tasks = task_step.get('all_tasks', {})
                print(f"\n🎯 Generated Tasks:")
                for task_type, tasks in all_tasks.items():
                    print(f"   {task_type.title()} Tasks ({len(tasks)}):")
                    for i, task in enumerate(tasks[:2]):  # 처음 2개만 표시
                        prompt = task.get('prompt', '')
                        print(f"     [{i+1}] {prompt[:80]}...")
            
            # 보상 분포 상세
            if 'reward_computation' in result['steps']:
                reward_step = result['steps']['reward_computation']
                rewards = reward_step.get('rewards', {})
                distribution = rewards.get('reward_distribution', {})
                print(f"\n🏆 Reward Distribution:")
                for task_type, avg_reward in distribution.items():
                    print(f"   {task_type.title()}: {avg_reward:.3f}")
        
        print("\n" + "="*60)
        print("🎉 PIPELINE TEST COMPLETED SUCCESSFULLY")
        print("="*60)
        
        # 상세 결과 파일 저장
        if result['success']:
            print(f"\n📁 상세 결과 파일 저장 중...")
            save_detailed_results(result, args, args.output_dir)
        
        return result['success']
        
    except Exception as e:
        logger.log_error(f"💥 Pipeline test failed: {e}")
        import traceback
        traceback.print_exc()
        return False
    
    finally:
        # GPU 메모리 정리
        if torch.cuda.is_available():
            torch.cuda.empty_cache()
        logger.log_info("🧹 Cleaned up resources")


if __name__ == '__main__':
    success = main()
    exit_code = 0 if success else 1
    print(f"\n🚪 Exiting with code {exit_code}")
    sys.exit(exit_code)