diff --git a/paddle/fluid/platform/profiler.cc b/paddle/fluid/platform/profiler.cc
index 040a68f67264d0068fc484b6629828bd2bdd222a..85977366e61c676fc5d2d3c5d22dd2f606543684 100644
--- a/paddle/fluid/platform/profiler.cc
+++ b/paddle/fluid/platform/profiler.cc
@@ -186,7 +186,6 @@ RecordEvent::RecordEvent(const std::string& name, const DeviceContext* dev_ctx)
 
 RecordEvent::~RecordEvent() {
   if (g_state == ProfilerState::kDisabled || !is_enabled_) return;
-  VLOG(5) << "call ~RecordEvent";
   std::lock_guard<std::mutex> l(profiler_mu);
   DeviceTracer* tracer = GetDeviceTracer();
   if (tracer) {
diff --git a/python/paddle/fluid/tests/unittests/test_parallel_executor_mnist.py b/python/paddle/fluid/tests/unittests/test_parallel_executor_mnist.py
index 0ff7b73123b59a2f9254c88662fa7610b2d7b7e8..63bc1de208d198c5d55b034ed6ce2af184449bc9 100644
--- a/python/paddle/fluid/tests/unittests/test_parallel_executor_mnist.py
+++ b/python/paddle/fluid/tests/unittests/test_parallel_executor_mnist.py
@@ -138,6 +138,13 @@ class TestMNIST(TestParallelExecutorBase):
                        "label": label},
             use_cuda=use_cuda,
             use_parallel_executor=False)
+        parallel_first_loss, parallel_last_loss = self.check_network_convergence(
+            method=simple_fc_net,
+            seed=1,
+            feed_dict={"image": img,
+                       "label": label},
+            use_cuda=use_cuda,
+            use_parallel_executor=True)
 
         self.assertAlmostEquals(
             np.mean(parallel_first_loss),
@@ -149,8 +156,6 @@ class TestMNIST(TestParallelExecutorBase):
     def test_simple_fc_parallel_accuracy(self):
         if core.is_compiled_with_cuda():
             self.check_simple_fc_parallel_accuracy(True)
-            self.check_simple_fc_parallel_accuracy(True)
-        # FIXME(Yancey1989): ParallelGraph executor type support CPU mode
         self.check_simple_fc_parallel_accuracy(False)
 
     def check_batchnorm_fc_convergence(self, use_cuda, use_fast_executor):
@@ -171,9 +176,6 @@ class TestMNIST(TestParallelExecutorBase):
             for use_fast_executor in (False, True):
                 self.check_batchnorm_fc_convergence(use_cuda, use_fast_executor)
 
-        self.check_batchnorm_fc_convergence(
-            use_cuda=True, use_fast_executor=False)
-
     def test_batchnorm_fc_with_new_strategy(self):
         # FIXME(zcd): close this test temporally.
         # self._compare_reduce_and_allreduce(fc_with_batchnorm, True)
diff --git a/python/paddle/fluid/tests/unittests/test_parallel_executor_seresnext.py b/python/paddle/fluid/tests/unittests/test_parallel_executor_seresnext.py
index 4f1d902f5c380fb5141deffda8f431c1458eff0f..e7a56bb6386a812e43e5c1b5c08cd0682aa9223a 100644
--- a/python/paddle/fluid/tests/unittests/test_parallel_executor_seresnext.py
+++ b/python/paddle/fluid/tests/unittests/test_parallel_executor_seresnext.py
@@ -313,9 +313,7 @@ class TestResnet(TestParallelExecutorBase):
             np.mean(parallel_last_loss), single_last_loss[0], delta=delta2)
 
     def test_seresnext_with_learning_rate_decay(self):
-        if core.is_compiled_with_cuda():
-            self._check_resnet_convergence(
-                model=SE_ResNeXt50Small, use_cuda=True)
+        self._check_resnet_convergence(model=SE_ResNeXt50Small, use_cuda=True)
         self._check_resnet_convergence(
             model=SE_ResNeXt50Small, use_cuda=False, iter=2, delta2=1e-3)