diff --git a/demo/mnist/api_train_v2.py b/demo/mnist/api_train_v2.py
index 06beb7024d1fd07dc327cb4c09d74e1b89a7b8ff..575a32b3222a83afa0f01bef59037dca102773e7 100644
--- a/demo/mnist/api_train_v2.py
+++ b/demo/mnist/api_train_v2.py
@@ -30,10 +30,11 @@ def main():
                 result = trainer.test(reader=paddle.reader.batched(
                     paddle.dataset.mnist.test(), batch_size=256))
 
-                print "Pass %d, Batch %d, Cost %f, %s, Testing metrics %s" % (
-                    event.pass_id, event.batch_id, event.cost, event.metrics,
-                    result.metrics)
-
+                print "Pass %d, Batch %d, Cost %.2f, %s\n" \
+                      "Testing cost %.2f metrics %s" % (
+                          event.pass_id, event.batch_id, event.cost,
+                          event.metrics,
+                          result.cost, result.metrics)
         else:
             pass
 
diff --git a/python/paddle/v2/event.py b/python/paddle/v2/event.py
index a78bcf076cc65e0dfdfc5760e099900418162f35..a429e36b63c9e812332673b66f4d8b99f3303cf8 100644
--- a/python/paddle/v2/event.py
+++ b/python/paddle/v2/event.py
@@ -34,8 +34,9 @@ class WithMetric(object):
 
 
 class TestResult(WithMetric):
-    def __init__(self, evaluator):
+    def __init__(self, evaluator, cost):
         super(TestResult, self).__init__(evaluator)
+        self.cost = cost
 
 
 class BeginPass(object):
diff --git a/python/paddle/v2/trainer.py b/python/paddle/v2/trainer.py
index abaad1d0213c63f92a72b30158d41fd41119d078..b4a713f7d53c2c5aef7e356906f88475037fa8d2 100644
--- a/python/paddle/v2/trainer.py
+++ b/python/paddle/v2/trainer.py
@@ -120,10 +120,8 @@ class SGD(ITrainer):
                 for each_param in self.__gradient_machine__.getNonStaticParameters(
                 ):
                     updater.update(each_param)
-                # Get cost. We use numpy to calculate total cost for this batch.
-                cost_vec = out_args.getSlotValue(0)
-                cost_vec = cost_vec.copyToNumpyMat()
-                cost = cost_vec.sum() / len(data_batch)
+                cost_sum = out_args.sumCosts()
+                cost = cost_sum / len(data_batch)
                 updater.finishBatch(cost)
                 batch_evaluator.finish()
                 event_handler(
@@ -152,13 +150,18 @@ class SGD(ITrainer):
         evaluator = self.__gradient_machine__.makeEvaluator()
         out_args = api.Arguments.createArguments(0)
         evaluator.start()
+        total_cost = 0
+        num_samples = 0.0
         for data_batch in reader():
+            num_samples += len(data_batch)
             self.__gradient_machine__.forward(
                 feeder(data_batch), out_args, api.PASS_TEST)
+            total_cost += out_args.sumCosts()
             self.__gradient_machine__.eval(evaluator)
 
         evaluator.finish()
-        return v2_event.TestResult(evaluator=evaluator)
+        return v2_event.TestResult(
+            evaluator=evaluator, cost=total_cost / num_samples)
 
 
 def __check_train_args__(reader, event_handler, **kwargs):