measure_weights: gold measure of any weights into the registry (the gate's baseline)

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

.gitignore

@@ -437,3 +437,4 @@ runs/*/ultralytics/
 runs/*/gold_predictions/
 runs/*/dataset.resolved.yaml
 runs/slurm-*.out
+runs/_measure_*/

tools/measure_weights.py

@@ -0,0 +1,53 @@
+#!/usr/bin/env python3
+"""Measure any weights on the gold set and store the report in the registry.
+
+Used once to give the deployed model its ``gold_report.json``, the baseline the
+gate compares every new run against:
+
+    python tools/measure_weights.py /path/to/best.pt --version yolo26n-seg-overlap-false-2026-04-12
+"""
+
+from __future__ import annotations
+
+import argparse
+import json
+import sys
+from datetime import datetime
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parents[1]
+sys.path.insert(0, str(ROOT))
+
+from tools.train_run import measure  # noqa: E402
+
+DEFAULT_NAMES = ["loop_all", "pin", "crystal", "loop_face", "ice", "needle"]
+
+
+def main(argv=None) -> int:
+    ap = argparse.ArgumentParser(description=__doc__, formatter_class=argparse.RawDescriptionHelpFormatter)
+    ap.add_argument("weights")
+    ap.add_argument("--version", required=True, help="registry directory name to write gold_report.json into")
+    ap.add_argument("--conf", type=float, default=0.25)
+    ap.add_argument("--names", default=",".join(DEFAULT_NAMES))
+    args = ap.parse_args(argv)
+    reg = ROOT / "models" / "registry" / args.version
+    reg.mkdir(parents=True, exist_ok=True)
+    work = ROOT / "runs" / f"_measure_{args.version}"
+    work.mkdir(parents=True, exist_ok=True)
+    report = measure(Path(args.weights), work, args.names.split(","), conf=args.conf)
+    report["weights"] = str(Path(args.weights).resolve())
+    report["measured_at"] = datetime.now().isoformat(timespec="seconds")
+    out = reg / "gold_report.json"
+    out.write_text(json.dumps(report, indent=2) + "\n", encoding="utf-8")
+    meta = reg / "metadata.json"
+    if not meta.exists():
+        meta.write_text(json.dumps({"model_id": args.version, "task": "segmentation",
+                                    "metrics": report["overall"], "note": "metadata written by measure_weights.py"},
+                                   indent=2) + "\n", encoding="utf-8")
+    o = report["overall"]
+    print(f"{args.version}: f1 {o['f1']:.3f} mask IoU {o['mean_mask_iou']:.3f} -> {out}")
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())