diff --git a/inplace_abn/abn.py b/inplace_abn/abn.py
index caab17e..e778dea 100644
--- a/inplace_abn/abn.py
+++ b/inplace_abn/abn.py
@@ -4,6 +4,7 @@
 import torch.nn.functional as functional
 
 from .functions import *
+_default_group = distributed.group.WORLD if hasattr(distributed, "group") else None
 
 
 class ABN(nn.Module):
@@ -138,7 +139,7 @@ class InPlaceABNSync(ABN):
     """
 
     def __init__(self, num_features, eps=1e-5, momentum=0.1, affine=True, activation="leaky_relu",
-                 activation_param=0.01, group=distributed.group.WORLD):
+                 activation_param=0.01, group=_default_group):
         super(InPlaceABNSync, self).__init__(num_features, eps, momentum, affine, activation, activation_param)
         self.group = group
 
diff --git a/inplace_abn/functions.py b/inplace_abn/functions.py
index d2cb4c5..51ccf81 100644
--- a/inplace_abn/functions.py
+++ b/inplace_abn/functions.py
@@ -3,6 +3,7 @@
 from torch.autograd.function import once_differentiable
 
 from . import _backend
+_default_group = distributed.group.WORLD if hasattr(distributed, "group") else None
 
 
 def _activation_from_name(activation):
@@ -152,7 +153,7 @@ def inplace_abn(x, weight, bias, running_mean, running_var,
 
 def inplace_abn_sync(x, weight, bias, running_mean, running_var,
                      training=True, momentum=0.1, eps=1e-05, activation="leaky_relu", activation_param=0.01,
-                     group=distributed.group.WORLD):
+                     group=_default_group):
     return InPlaceABN.apply(x, weight, bias, running_mean, running_var,
                             training, momentum, eps, activation, activation_param, group)