Wire W&B tracking into ISMCTS trainer

src/coolrl_lost_cities/games/classic/ismcts/cli.py

@@ -8,6 +8,8 @@ from typing import Any
 
 import yaml
 
+from coolrl_lost_cities.games.classic.deep_cfr.tracking import WandbRunTracker
+
 from .config import IsMctsConfig, load_config
 from .trainer import IsMctsTrainer
 
@@ -57,13 +59,31 @@ def train_command(args: argparse.Namespace) -> None:
     config = load_config(args.config) if args.config else IsMctsConfig()
     config = _with_overrides(config, args.config_overrides)
     run_dir = _resolve_run_dir(config, keep=args.keep)
+    tracker = None
+    if args.wandb:
+        tracker = WandbRunTracker(
+            project=args.wandb_project,
+            name=args.wandb_name or config.run.experiment_name,
+            mode=args.wandb_mode,
+            run_dir=run_dir,
+            config=config.to_dict() if hasattr(config, "to_dict") else config.model_dump(),
+            group=args.wandb_group,
+            job_type=args.wandb_job_type,
+            tags=list(args.wandb_tag) if args.wandb_tag else None,
+            notes=args.wandb_notes,
+        )
     trainer = IsMctsTrainer(
         config,
         config.rules.to_lost_cities_config(seed=config.run.seed),
         run_dir=run_dir,
         device=config.run.device,
+        tracker=tracker,
     )
-    trainer.train()
+    try:
+        trainer.train()
+    finally:
+        if tracker is not None:
+            tracker.close()
 
 
 def main(argv: list[str] | None = None) -> None:
@@ -79,7 +99,11 @@ def main(argv: list[str] | None = None) -> None:
         dest="config_overrides",
         metavar="PATH=VALUE",
     )
-    train.add_argument("--wandb", action="store_true", help="Accepted for CLI parity; ignored.")
+    train.add_argument(
+        "--wandb",
+        action="store_true",
+        help="Mirror metrics to Weights & Biases (requires wandb extra).",
+    )
     train.add_argument("--wandb-project", default="coolrl-lost-cities")
     train.add_argument("--wandb-name")
     train.add_argument("--wandb-mode", choices=("online", "offline", "disabled"), default="online")

src/coolrl_lost_cities/games/classic/ismcts/trainer.py

@@ -57,11 +57,13 @@ class IsMctsTrainer:
         *,
         run_dir: str | Path,
         device: torch.device | str = "cpu",
+        tracker: object | None = None,
     ) -> None:
         self.config = config
         self.game_config = game_config
         self.run_dir = Path(run_dir)
         self.device = self._resolve_device(device)
+        self.tracker = tracker
         probe = GameState.new_game(game_config, seed=config.run.seed)
         self.input_dim = input_dim(probe, config.encoding)
         self.action_size = probe.action_size
@@ -101,6 +103,11 @@ class IsMctsTrainer:
             metrics.append(item)
             self._append_metrics(item)
             self._save_checkpoints(iteration, item)
+            if self.tracker is not None:
+                try:
+                    self.tracker.log_metrics(item.to_dict(), step=iteration)
+                except Exception as exc:  # pragma: no cover
+                    print(f"tracker.log_metrics failed: {exc}", flush=True)
             print(json.dumps(item.to_dict(), sort_keys=True))
         return metrics