scripts/report.py

#!/usr/bin/env python3
"""
Aggregate multi-run evaluation report for SYNAPSE-X.
"""

import json
import sys
from pathlib import Path
from statistics import mean, stdev

PROJECT_ROOT = Path(__file__).resolve().parents[1]
if str(PROJECT_ROOT) not in sys.path:
    sys.path.insert(0, str(PROJECT_ROOT))

from scripts.inference import run_episode


def evaluate(level: str, runs: int = 20, start_seed: int = 42):
    scores = [run_episode(level, seed=start_seed + idx, verbose=False) for idx in range(runs)]
    return {
        "mean": round(mean(scores), 4),
        "std": round(stdev(scores) if len(scores) > 1 else 0.0, 4),
        "min": round(min(scores), 4),
        "max": round(max(scores), 4),
    }


def main():
    report = {}
    for level in ["easy", "medium", "hard"]:
        stats = evaluate(level, runs=20)
        report[level] = stats
        print(f"{level}: {stats}")

    with open(PROJECT_ROOT / "report_results.json", "w", encoding="utf-8") as handle:
        json.dump(report, handle, indent=2)


if __name__ == "__main__":
    main()