fix local variable override bug (fix #430)

tensorpack/train/distributed.py

@@ -13,7 +13,6 @@ from ..tfutils.sesscreate import NewSessionCreator
 from ..tfutils.common import get_global_step_var, get_op_tensor_name
 
 from .multigpu import MultiGPUTrainerBase
-from .utility import override_to_local_variable
 
 
 __all__ = ['DistributedTrainerReplicated']
@@ -205,15 +204,14 @@ class DistributedTrainerReplicated(MultiGPUTrainerBase):
         cbs = self._input_source.setup(self.model.get_inputs_desc())
         self.config.callbacks.extend(cbs)
 
-        with override_to_local_variable():
-            # Ngpu * Nvar * 2
-            grad_list = MultiGPUTrainerBase.build_on_multi_tower(
-                self.config.tower,
-                lambda: MultiGPUTrainerBase._build_graph_get_grads(
-                    self.model, self._input_source),
-                devices=self.raw_devices,
-                use_vs=[True] * self.config.nr_tower)  # open vs at each tower
-            MultiGPUTrainerBase._check_grad_list(grad_list)
+        # Ngpu * Nvar * 2
+        grad_list = MultiGPUTrainerBase.build_on_multi_tower(
+            self.config.tower,
+            lambda: MultiGPUTrainerBase._build_graph_get_grads(
+                self.model, self._input_source),
+            devices=self.raw_devices,
+            use_vs=[True] * self.config.nr_tower)  # open vs at each tower
+        MultiGPUTrainerBase._check_grad_list(grad_list)
 
         avg_grads = DistributedTrainerReplicated._average_grads(grad_list, self.raw_devices)
         with tf.device(self.param_server_device):