Adjust AdversarialBalancing to numpy>=2

replace `row_stack` with `vstack` and `np.NaN` with `np.nan`.

Signed-off-by: Ehud-Karavani <[email protected]>

causallib/contrib/adversarial_balancing/adversarial_balancing.py

@@ -137,7 +137,7 @@ def _run(self, X, A, w_init=None, is_train=True, use_stabilized=None, **select_k
             if not np.all(unique_treatments == np.arange(n_treatments)):
                 raise AssertionError("Treatment values in `a` must be indexed 0, 1, 2, ...")
             self.iterative_models_ = np.empty((n_treatments, self.iterations), dtype=object)
-            self.iterative_normalizing_consts_ = np.full((n_treatments, self.iterations), np.NaN)
+            self.iterative_normalizing_consts_ = np.full((n_treatments, self.iterations), np.nan)
 
             self.discriminator_loss_ = np.zeros((n_treatments, self.iterations))
             self.treatments_frequency_ = _compute_treatments_frequency(A)
@@ -147,7 +147,7 @@ def _run(self, X, A, w_init=None, is_train=True, use_stabilized=None, **select_k
             # population ("source population"),
             # and the samples with label -1 are the population under treatment a ("target population").
             # Labels 1 and -1 (rather than 0) are used because of the later exponential loss function
-            X_augm = np.row_stack((X, X[A == a]))  # create the augmented dataset
+            X_augm = np.vstack((X, X[A == a]))  # create the augmented dataset
             y = np.ones((X_augm.shape[0]))
             y[X.shape[0]:] *= -1  # subpopulation of current treatment (a) has y== -1
             target_pop_mask = y == -1

causallib/contrib/tests/test_adversarial_balancing.py

@@ -16,7 +16,7 @@ class TestAdversarialBalancing(unittest.TestCase):
     def create_identical_treatment_groups_data(n=100):
         np.random.seed(42)
         X = np.random.rand(n, 3)
-        X = np.row_stack((X, X))  # Duplicate identical samples
+        X = np.vstack((X, X))  # Duplicate identical samples
         a = np.array([1] * n + [0] * n)  # Give duplicated samples different treatment assignment
         X, a = pd.DataFrame(X), pd.Series(a)
         return X, a