debug_consensus.py

#!/usr/bin/env python3
"""
Debug consensus mechanism.
"""

import sys
from pathlib import Path

# Add src to path
sys.path.append(str(Path(__file__).parent / "src"))

from models.model_factory import ModelFactory
from adaptive.adaptive_cot import AdaptiveCoT

def debug_consensus():
    """Debug the consensus mechanism."""
    print("🔬 Debug Consensus Mechanism")
    print("=" * 60)
    
    # Load model
    print("🔧 Loading model...")
    model = ModelFactory.create_model(
        model_type="deepseek",
        model_name="/raid/LLM/deepseek-r1-distill-qwen-1.5b",
        config={"gpu_id": 0}
    )
    model.load_model()
    
    # Create adaptive CoT
    config = {
        "adaptive_branching": False,
        "min_branches": 8,
        "max_branches": 8,
        "default_branches": 8,
        "num_fewshot": 0,
        "temperature": 0.7,
        "top_p": 0.95,
        "max_tokens": 512,
    }
    
    adaptive_cot = AdaptiveCoT(model, config)
    
    # Test with the house flipping problem
    problem = "Josh decides to try flipping a house.  He buys a house for $80,000 and then puts in $50,000 in repairs.  This increased the value of the house by 150%.  How much profit did he make?"
    
    print(f"📝 Problem: {problem}")
    print("=" * 60)
    
    # Solve the problem
    result = adaptive_cot.solve_problem(problem)
    
    print(f"🎯 Final answer: {result['final_answer']}")
    print(f"📊 Consensus confidence: {result['consensus_info']['confidence']:.3f}")
    print(f"📊 Answer counts: {result['consensus_info']['answer_counts']}")
    print()
    
    # Show individual branch answers
    print("🔍 Individual branch answers:")
    for i, (reasoning, answer) in enumerate(zip(result['reasoning_paths'], result['extracted_answers'])):
        print(f"Branch {i+1}: {answer}")
        # Show first 100 chars of reasoning
        reasoning_preview = reasoning[:100].replace('\n', ' ')
        print(f"  Reasoning: {reasoning_preview}...")
        print()

if __name__ == "__main__":
    debug_consensus()