Internal fix for NumPy 1.24 test failures.

PiperOrigin-RevId: 503544399 Change-Id: Icf9acb48a940870a8479c89b3850f3dc01d49382
KezhiAdore · Jan 23, 2023 · bbe0007 · bbe0007
1 parent 87b4cce
commit bbe0007
Show file tree

Hide file tree

Showing 3 changed files with 4 additions and 3 deletions.
diff --git a/open_spiel/python/examples/meta_cfr/sequential_games/dataset_generator.py b/open_spiel/python/examples/meta_cfr/sequential_games/dataset_generator.py
@@ -27,7 +27,7 @@ class Dataset:
   def __init__(self, train_dataset: List[Tuple[List[List[float]],
                                                InfostateNode]],
                batch_size: int):
-    self._train_dataset = np.array(train_dataset)
+    self._train_dataset = np.array(train_dataset, dtype=object)
     self._size = self._train_dataset.shape[0]
     self._batch_size = batch_size
 

diff --git a/open_spiel/python/examples/meta_cfr/sequential_games/meta_learning.py b/open_spiel/python/examples/meta_cfr/sequential_games/meta_learning.py
@@ -445,7 +445,7 @@ def training_optimizer(self):
         for _ in range(FLAGS.num_batches):
           batch = next(data_loader)
           cfvalues, infoset = zip(*batch)
-          cfvalues = np.array(list(cfvalues))
+          cfvalues = np.array(list(cfvalues), dtype=object)
           cfvalues = utils.mask(cfvalues, infoset, len(self._all_actions),
                                 FLAGS.batch_size)
           self.optimize_infoset(cfvalues, infoset, self._infostate_map,

diff --git a/open_spiel/python/examples/meta_cfr/sequential_games/utils.py b/open_spiel/python/examples/meta_cfr/sequential_games/utils.py
@@ -47,7 +47,8 @@ def get_batched_input(input_list: List[jax.numpy.DeviceArray],
                                   1) - len(input_list)
   idx_sample = np.random.choice(len(input_list), items_to_sample)
   input_zip = np.array(
-      list(zip(input_list, infostate_list, illegal_action_list)))
+      list(zip(input_list, infostate_list, illegal_action_list)),
+      dtype=object)
   input_lst_sample = input_zip[idx_sample]
   input_sample, infostate_sample, illegal_action_sample = zip(*input_lst_sample)