Continue rejected v37 through frozen loop

scripts/run_belgium_building_training_loop.py

@@ -91,6 +91,7 @@ def training_command(
     project: Path,
     name: str,
     epochs: int,
+    patience: int = 18,
     seed: int,
     batch: int,
     workers: int,
@@ -115,7 +116,7 @@ def training_command(
         f"batch={batch}",
         "device=0",
         f"workers={workers}",
-        "patience=35",
+        f"patience={patience}",
         "cache=disk",
         "close_mosaic=20",
         f"max_det={max_det}",
@@ -180,6 +181,7 @@ def main() -> int:
     parser.add_argument("--output-dir", type=Path, required=True)
     parser.add_argument("--iterations", type=int, default=1)
     parser.add_argument("--epochs", type=int, default=160)
+    parser.add_argument("--patience", type=int, default=18)
     parser.add_argument("--batch", type=int, default=2)
     parser.add_argument("--workers", type=int, default=4)
     parser.add_argument("--max-det", type=int, default=1000)
@@ -236,6 +238,7 @@ def main() -> int:
         index = first_index + offset
         name = f"iteration-{index:03d}"
         iteration_dir = args.output_dir / name
+        iteration_dir.mkdir(parents=True, exist_ok=True)
         train_run = args.output_dir / "runs" / name
         evaluate_existing = args.evaluate_initial_model and offset == 0 and not state["iterations"]
         command = None if evaluate_existing else training_command(
@@ -244,7 +247,8 @@ def main() -> int:
             data=train_yaml,
             project=args.output_dir / "runs",
             name=name,
-            epochs=args.epochs,
+                epochs=args.epochs,
+                patience=args.patience,
             seed=args.seed + index,
             batch=args.batch,
             workers=args.workers,