Make S/R matrix runner cross-platform

2026-07-12 21:27:13 +02:00
parent 19b81c169d
commit 681f0f95da
4 changed files with 144 additions and 70 deletions
@@ -0,0 +1,128 @@
+"""Run the S/R v2 training matrix or one locked validation comparison.
+
+This is a cross-platform orchestrator around ``run_backtest_snapshot.py``. It
+contains no backtest logic; every arm still runs through the production-parity
+Python harness.
+"""
+
+from __future__ import annotations
+
+import argparse
+import subprocess
+import sys
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parents[1]
+RUNNER = ROOT / "scripts" / "run_backtest_snapshot.py"
+COMPARE = ROOT / "scripts" / "compare_sr_variants.py"
+TRAINING_ARMS = (
+    "production_control",
+    "rr_aligned_control",
+    "rewrite",
+    "soft_zones",
+    "confirmed_rounds",
+    "gate_v2",
+)
+LOCKABLE_ARMS = TRAINING_ARMS[1:]
+
+
+def _add_common(parser: argparse.ArgumentParser) -> None:
+    parser.add_argument(
+        "--snapshot",
+        default="backtest_snapshots/prod.sqlite",
+        help="Local SQLite snapshot path.",
+    )
+    parser.add_argument("--workers", type=int, default=7)
+
+
+def _args() -> argparse.Namespace:
+    parser = argparse.ArgumentParser(description=__doc__)
+    commands = parser.add_subparsers(dest="command", required=True)
+    train = commands.add_parser("train", help="Run all arms before 2024-07-01.")
+    _add_common(train)
+    validate = commands.add_parser(
+        "validate",
+        help="Run production control and one locked arm from 2024-07-01.",
+    )
+    _add_common(validate)
+    validate.add_argument("--locked-arm", required=True, choices=LOCKABLE_ARMS)
+    return parser.parse_args()
+
+
+def _run_arm(
+    arm: str,
+    snapshot: str,
+    workers: int,
+    *,
+    entry_flag: str,
+    entry_date: str,
+    output: Path,
+) -> None:
+    print(f"Running S/R arm: {arm}", flush=True)
+    subprocess.run(
+        [
+            sys.executable,
+            str(RUNNER),
+            snapshot,
+            "--workers", str(workers),
+            "--allow-spawn",
+            "--sr-variant", arm,
+            entry_flag, entry_date,
+            "--sr-audit",
+            "--out", str(output),
+        ],
+        cwd=ROOT,
+        check=True,
+    )
+
+
+def _train(args: argparse.Namespace) -> None:
+    for arm in TRAINING_ARMS:
+        _run_arm(
+            arm,
+            args.snapshot,
+            args.workers,
+            entry_flag="--entry-end",
+            entry_date="2024-06-30",
+            output=ROOT / "reports" / f"backtest-sr-v2-train-{arm}.json",
+        )
+    print("Training matrix complete. Lock one arm before running validation.")
+
+
+def _validate(args: argparse.Namespace) -> None:
+    reports: dict[str, Path] = {}
+    for arm in ("production_control", args.locked_arm):
+        output = ROOT / "reports" / f"backtest-sr-v2-validation-{arm}.json"
+        reports[arm] = output
+        _run_arm(
+            arm,
+            args.snapshot,
+            args.workers,
+            entry_flag="--entry-start",
+            entry_date="2024-07-01",
+            output=output,
+        )
+    subprocess.run(
+        [
+            sys.executable,
+            str(COMPARE),
+            str(reports["production_control"]),
+            str(reports[args.locked_arm]),
+            "--out-csv", str(ROOT / "reports" / "sr-v2-validation-cohorts.csv"),
+            "--out-json", str(ROOT / "reports" / "sr-v2-validation-comparison.json"),
+        ],
+        cwd=ROOT,
+        check=True,
+    )
+
+
+def main() -> None:
+    args = _args()
+    if args.command == "train":
+        _train(args)
+    else:
+        _validate(args)
+
+
+if __name__ == "__main__":
+    main()