Program Performance Evaluation

This section covers how to use the Evaluator to assess generated code for optimization tasks.

Modules

• src.evaluator.Evaluator: Main class for evaluating code snippets.
• execute_single_seed: Helper for single-run metrics.
• compute_total_distance: Example utility for TSP tasks.

Usage Example

from src.database import Database, DatabaseConfig
from src.evaluator import Evaluatorrom src.tasks.tsp_task import TSPTask

# Load database of evolved programs
db = Database(DatabaseConfig(database_path='evolution.db'))
entries = db.get_all_programs()

evaluator = Evaluator()
task = TSPTask(n_cities=10)

for entry in entries:
    code = entry.code
    gen = entry.generation_number
    result = evaluator.evaluate_code(code, task, seeds=[0,1,2])
    print(f"Generation {gen}: cost={result.cost:.2f}, feasible={result.feasible}")

Configuration

Adjust timeouts and memory limits via Evaluator(timeout=60.0, max_memory_mb=256).