diff --git a/core/metric.py b/core/metric.py
index 287ec2f8a2007739eaa3303933a345ea10b063fb..d9968fa40167b6ca728b0c1046fca5e70ef427a7 100755
--- a/core/metric.py
+++ b/core/metric.py
@@ -13,6 +13,8 @@
 # limitations under the License.
 
 import abc
+import paddle.fluid as fluid
+import numpy as np
 
 
 class Metric(object):
@@ -24,14 +26,23 @@ class Metric(object):
         """ """
         pass
 
-    def clear(self, scope, params):
+    def clear(self, scope=None, **kwargs):
         """
         clear current value
         Args:
             scope: value container
             params: extend varilable for clear
         """
-        pass
+        if scope is None:
+            scope = fluid.global_scope()
+
+        place = fluid.CPUPlace()
+        for (varname, dtype) in self._need_clear_list:
+            if scope.find_var(varname) is None:
+                continue
+            var = scope.var(varname).get_tensor()
+            data_array = np.zeros(var._get_dims()).astype(dtype)
+            var.set(data_array, place)
 
     def calculate(self, scope, params):
         """
diff --git a/core/metrics/precision.py b/core/metrics/precision.py
index d748e17d3ee6c34b282ee3558d35ce6d4f66c952..4b9b4bd3101854f70308455cabc67bb64249b5dc 100755
--- a/core/metrics/precision.py
+++ b/core/metrics/precision.py
@@ -99,6 +99,8 @@ class Precision(Metric):
 
         self.accuracy = local_pos_num / local_ins_num
 
+        self._need_clear_list = [("local_ins_num", "float32"),
+                                 ("local_pos_num", "float32")]
         self.metrics = dict()
         metric_varname = "P@%d" % kwargs.get("k")
         self.metrics[metric_varname] = self.accuracy
diff --git a/core/trainers/framework/runner.py b/core/trainers/framework/runner.py
index f316327ac2447d1a7ac40f511d487685d33e542b..d5fced11ffd546b36ee7db3e596f061bf8a58328 100644
--- a/core/trainers/framework/runner.py
+++ b/core/trainers/framework/runner.py
@@ -221,6 +221,7 @@ class RunnerBase(object):
         program = context["model"][model_name]["main_program"].clone()
         _exe_strategy, _build_strategy = self._get_strategy(model_dict,
                                                             context)
+
         program = fluid.compiler.CompiledProgram(program).with_data_parallel(
             loss_name=model_class.get_avg_cost().name,
             build_strategy=_build_strategy,
@@ -497,6 +498,10 @@ class SingleInferRunner(RunnerBase):
             with fluid.program_guard(train_prog, startup_prog):
                 fluid.io.load_persistables(
                     context["exe"], model_path, main_program=train_prog)
+            clear_metrics = context["model"][model_dict["name"]][
+                "model"].get_clear_metrics()
+            for var in clear_metrics:
+                var.clear()
 
     def _dir_check(self, context):
         dirname = envs.get_global_env(