Auto-sync: 2026-06-28 07:13:10 (part 3)

tests/test_trainer.py

@@ -10,7 +10,9 @@ from dovla_cil.training.trainer import (
     DoVLATrainer,
     TrainerConfig,
     _best_records_by_group,
+    _coerce_policy_target_action,
     _cross_state_pair_indices,
+    _load_policy_target_action_map,
     _reward_utility_values,
 )
 from dovla_cil.utils.io import read_json
@@ -178,3 +180,26 @@ def test_policy_target_map_overrides_group_target_with_fallback() -> None:
         "g0": "field_choice",
         "g1": "fallback",
     }
+
+
+def test_policy_target_action_map_loads_continuous_targets(tmp_path: Path) -> None:
+    path = tmp_path / "targets.json"
+    path.write_text(
+        """
+        {
+          "targets": {
+            "g0": {"record_id": "r0", "action_values": [[0.1, 0.2], [0.3, 0.4]]},
+            "g1": "legacy_record_id"
+          }
+        }
+        """
+    )
+
+    loaded = _load_policy_target_action_map(path)
+
+    assert loaded == {"g0": [[0.1, 0.2], [0.3, 0.4]]}
+    assert _coerce_policy_target_action(
+        loaded["g0"],
+        action_dim=3,
+        action_horizon=3,
+    ) == [[0.1, 0.2, 0.0], [0.3, 0.4, 0.0], [0.0, 0.0, 0.0]]