Spaces:
Sleeping
Sleeping
File size: 4,144 Bytes
2ef845d | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 | from __future__ import annotations
import argparse
import json
from datetime import datetime, timezone
from pathlib import Path
from typing import Any, Dict
def parse_args() -> argparse.Namespace:
parser = argparse.ArgumentParser(description="Generate a hackathon-ready submission bundle from training artifacts")
parser.add_argument(
"--eval-summary",
default="artifacts/ppo_tuned_v3/evaluation_summary.json",
help="Path to evaluation_summary.json",
)
parser.add_argument(
"--output-dir",
default="artifacts/submission_bundle",
help="Directory to write submission bundle files",
)
parser.add_argument(
"--run-command",
default=(
"python scripts/train_sb3_agent.py --stage-steps 8000,8000,8000 "
"--entropy-schedule 0.02,0.01,0.003 --eval-freq 100000000 --output-dir artifacts/ppo_tuned_v3"
),
help="Training command used to produce the metrics",
)
return parser.parse_args()
def _load_json(path: Path) -> Dict[str, Any]:
if not path.exists():
raise FileNotFoundError(f"Missing file: {path}")
return json.loads(path.read_text(encoding="utf-8"))
def _task_row(task: str, summary: Dict[str, Any]) -> Dict[str, Any]:
payload = summary.get(task, {})
scores = payload.get("scores", [])
return {
"task": task,
"avg": float(payload.get("avg_score", 0.0) or 0.0),
"min": float(payload.get("min_score", 0.0) or 0.0),
"max": float(payload.get("max_score", 0.0) or 0.0),
"scores": [float(value) for value in scores],
}
def main() -> None:
args = parse_args()
project_root = Path(__file__).resolve().parents[1]
eval_summary_path = (project_root / args.eval_summary).resolve()
output_dir = (project_root / args.output_dir).resolve()
output_dir.mkdir(parents=True, exist_ok=True)
summary = _load_json(eval_summary_path)
tasks = ["clean_sweep", "tangled_lake", "legal_minefield", "cascade_collapse"]
rows = [_task_row(task, summary) for task in tasks]
legal_row = next(row for row in rows if row["task"] == "legal_minefield")
hard_gate = 0.75
bundle = {
"generated_at": datetime.now(timezone.utc).isoformat(),
"project": "gdpr-eraseops",
"training_command": args.run_command,
"evaluation_summary_path": str(eval_summary_path),
"tasks": rows,
"headline": {
"metric": "legal_minefield_avg",
"value": legal_row["avg"],
"target": hard_gate,
"pass": legal_row["avg"] >= hard_gate,
"margin": round(legal_row["avg"] - hard_gate, 4),
},
}
json_path = output_dir / "submission_report.json"
json_path.write_text(json.dumps(bundle, indent=2), encoding="utf-8")
md_lines = [
"# GDPR-EraseOps Submission Bundle",
"",
f"Generated at: {bundle['generated_at']}",
"",
"## Headline",
"",
f"- legal_minefield avg: {legal_row['avg']:.3f}",
f"- target: {hard_gate:.2f}",
f"- pass: {str(bundle['headline']['pass']).lower()}",
f"- margin: {bundle['headline']['margin']:+.3f}",
"",
"## Task Scores",
"",
"| Task | Avg | Min | Max | Seeds |",
"|---|---:|---:|---:|---|",
]
for row in rows:
seeds = ", ".join(f"{value:.2f}" for value in row["scores"])
md_lines.append(
f"| {row['task']} | {row['avg']:.3f} | {row['min']:.3f} | {row['max']:.3f} | [{seeds}] |"
)
md_lines.extend(
[
"",
"## Reproduce",
"",
"```bash",
args.run_command,
"```",
"",
"## Primary Sources",
"",
f"- {eval_summary_path}",
f"- {json_path}",
]
)
md_path = output_dir / "submission_report.md"
md_path.write_text("\n".join(md_lines) + "\n", encoding="utf-8")
print(json.dumps({"json": str(json_path), "markdown": str(md_path)}, indent=2))
if __name__ == "__main__":
main()
|