lint

open-thought · Mar 4, 2025 · bec36e6 · bec36e6
1 parent 9ce8ea6
commit bec36e6
Show file tree

Hide file tree

Showing 4 changed files with 13 additions and 12 deletions.
diff --git a/reasoning_gym/games/contrib/sokoban/src/generator.py b/reasoning_gym/games/contrib/sokoban/src/generator.py
@@ -85,14 +85,14 @@ def generate(
         if out_of_place_boxes >= boxes // 2:
             # Optionally save the puzzle to a file:
             if path:
-                np.savetxt(path, matrix, fmt='%s')
+                np.savetxt(path, matrix, fmt="%s")
             puzzle_str = player.puzzle_to_string(matrix)
 
             grid_list = [list(line) for line in puzzle_str.replace(" ", "").strip().split("\n")]
             grid_array = np.array(grid_list)
             solution, depth = solve_astar(grid_array, max_depth=max_depth)
             if solution is None:
-                continue    # retry generation
+                continue  # retry generation
 
             if debug:
                 print(f"solution={solution}")

diff --git a/reasoning_gym/graphs/family_relationships.py b/reasoning_gym/graphs/family_relationships.py
@@ -358,16 +358,14 @@ def _generate_story(self, family: set[Person]) -> str:
 
     def score_answer(self, answer: Optional[str], entry: dict[str, Any]) -> float:
         reward = 0.0
-        if answer is not None:
+        if isinstance(answer, str):
             try:
                 answer_formatted = answer.strip().lower()
-                solved = answer_formatted == entry["answer"].strip().lower()
-                if solved:
+                oracle_answer = entry["answer"].strip().lower()
+                if answer_formatted == oracle_answer:
                     reward = 1.0
-                else:
-                    reward = 0.01
             except:
-                reward = 0.01
+                pass
         return reward
 
 

diff --git a/tests_test_dataset_common.py → tests/test_dataset_common.py b/tests_test_dataset_common.py → tests/test_dataset_common.py
@@ -1,6 +1,7 @@
 import reasoning_gym
 from reasoning_gym.factory import DATASETS
 
+
 def test_score_answer_consistency():
     for dataset_name in DATASETS.keys():
         dataset = reasoning_gym.create_dataset(dataset_name, size=10, seed=1234)

diff --git a/tests/test_sokoban.py b/tests/test_sokoban.py
@@ -26,7 +26,9 @@ def test_sokoban():
         assert dataset.score_answer(answer=None, entry=item) == 0.0
 
     # Hard
-    config = SokobanConfig(seed=42, min_h=15, max_h=20, min_w=15, max_w=20, min_boxes=10, max_boxes=15, size=3, max_depth=90)
+    config = SokobanConfig(
+        seed=42, min_h=15, max_h=20, min_w=15, max_w=20, min_boxes=10, max_boxes=15, size=3, max_depth=90
+    )
     dataset = SokobanDataset(config)
 
     for item in dataset:
@@ -39,9 +41,10 @@ def test_sokoban():
         assert dataset.score_answer(answer=item["answer"], entry=item) == 1.0
         assert dataset.score_answer(answer=None, entry=item) == 0.0
 
-
     # min == max ranges
-    config = SokobanConfig(seed=42, min_h=11, max_h=11, min_w=11, max_w=11, min_boxes=11, max_boxes=11, size=3, max_depth=60)
+    config = SokobanConfig(
+        seed=42, min_h=11, max_h=11, min_w=11, max_w=11, min_boxes=11, max_boxes=11, size=3, max_depth=60
+    )
     dataset = SokobanDataset(config)
 
     for item in dataset:
@@ -53,4 +56,3 @@ def test_sokoban():
         # Test the scoring
         assert dataset.score_answer(answer=item["answer"], entry=item) == 1.0
         assert dataset.score_answer(answer=None, entry=item) == 0.0
-