Merge f02c6f8 into 2dce7e4

pynndescent/pynndescent_.py

@@ -803,10 +803,11 @@ def __init__(
 
         # Create a partial function for distances with arguments
         if len(self._dist_args) > 0:
+            dist_args = self._dist_args
 
             @numba.njit()
             def _partial_dist_func(x, y):
-                return _distance_func(x, y, *self._dist_args)
+                return _distance_func(x, y, *dist_args)
 
             self._distance_func = _partial_dist_func
         else:
@@ -874,12 +875,11 @@ def _partial_dist_func(x, y):
 
                     # Create a partial function for distances with arguments
                     if len(self._dist_args) > 0:
+                        dist_args = self._dist_args
 
                         @numba.njit()
                         def _partial_dist_func(ind1, data1, ind2, data2):
-                            return _distance_func(
-                                ind1, data1, ind2, data2, *self._dist_args
-                            )
+                            return _distance_func(ind1, data1, ind2, data2, *dist_args)
 
                         self._distance_func = _partial_dist_func
                     else:

pynndescent/sparse.py

@@ -16,7 +16,7 @@
 
 # Just reproduce a simpler version of numpy unique (not numba supported yet)
 @numba.njit()
-def isclose(a, b, rtol=1.e-5, atol=1.e-8):
+def isclose(a, b, rtol=1.0e-5, atol=1.0e-8):
     diff = np.abs(a - b)
     return diff <= (atol + rtol * np.abs(b))
 

pynndescent/tests/test_pynndescent_.py

@@ -331,8 +331,8 @@ def test_transformer_output_when_verbose_is_false():
     with redirect_stdout(out):
         _ = PyNNDescentTransformer(
             n_neighbors=4,
-            metric="euclidean",
-            metric_kwds={},
+            metric="standardised_euclidean",
+            metric_kwds={"sigma": np.ones(spatial_data.shape[1])},
             random_state=np.random,
             n_trees=5,
             n_iters=2,