diff --git a/paddle/fluid/framework/block_desc.h b/paddle/fluid/framework/block_desc.h
index ce48548418478cc5c9f9ca1244df9e66dca884e6..960ca39e1eadd3c064beb0e2c1342a406c4f0b6a 100644
--- a/paddle/fluid/framework/block_desc.h
+++ b/paddle/fluid/framework/block_desc.h
@@ -88,9 +88,8 @@ class BlockDesc {
   OpDesc *InsertOp(size_t index);
 
   /*
-   * Remove Op and its input/output variables.
-   * Note that for either input or output variable, if it is also an input or
-   * output variable of other ops, we should remain it.
+   * Only remove op itself,
+   * do nothing to its input and output variables
    */
   void RemoveOp(size_t s, size_t e);
 
diff --git a/paddle/fluid/platform/cudnn_helper.h b/paddle/fluid/platform/cudnn_helper.h
index 6ea4f8b7cba18ce7f803dbd9b15a7ae70c3055f2..bb8b14bb9fa41942c3aa653ca224c0842fbf9a00 100644
--- a/paddle/fluid/platform/cudnn_helper.h
+++ b/paddle/fluid/platform/cudnn_helper.h
@@ -59,13 +59,12 @@ inline const char* cudnnGetErrorString(cudnnStatus_t status) {
 #define CUDNN_VERSION_MIN(major, minor, patch) \
   (CUDNN_VERSION >= ((major)*1000 + (minor)*100 + (patch)))
 
-#define CUDNN_ENFORCE(condition)                                  \
-  do {                                                            \
-    cudnnStatus_t status = condition;                             \
-    if (status != CUDNN_STATUS_SUCCESS) {                         \
-      VLOG(1) << ::paddle::platform::cudnnGetErrorString(status); \
-      PADDLE_THROW("cuDNN call failed");                          \
-    }                                                             \
+#define CUDNN_ENFORCE(condition)                                     \
+  do {                                                               \
+    cudnnStatus_t status = condition;                                \
+    if (UNLIKELY(status != CUDNN_STATUS_SUCCESS)) {                  \
+      PADDLE_THROW(::paddle::platform::cudnnGetErrorString(status)); \
+    }                                                                \
   } while (false)
 
 enum class DataLayout {  // Not use
diff --git a/python/paddle/fluid/framework.py b/python/paddle/fluid/framework.py
index 53658610e55f6433dafe07547dc4f0fbb49493e2..6dc979dd5b0b5a348f6f6ea3da2259844d5a1602 100644
--- a/python/paddle/fluid/framework.py
+++ b/python/paddle/fluid/framework.py
@@ -1555,7 +1555,12 @@ class Program(object):
 
     def inference_optimize(self):
         """
-        This method will create a new program and change the :code:`is_test`
+        This method will create a new program and do following adjustments on it:
+        1. Remove all reader variables and their creator ops if exist.
+
+        2. Remove the :code:`read_op` if exists.
+
+        3. change the :code:`is_test`
         attribute of operators to :code:`True`. All the :code:`Parameter`
         information will be lost.
 
@@ -1569,6 +1574,22 @@ class Program(object):
         # core.inference_optimize being fixed.
         res = Program()
         res.desc = core.ProgramDesc(self.desc)
+
+        # remove all readers and the read_op if exist
+        read_op_idx = 0
+        root_block = res.desc.block(0)
+        while True:
+            if read_op_idx >= root_block.op_size() or root_block.op(
+                    read_op_idx).type() == 'read':
+                break
+            read_op_idx += 1
+        if read_op_idx < root_block.op_size():
+            root_block._remove_op(0, read_op_idx + 1)
+        for var in root_block.all_vars():
+            if var.type() == core.VarDesc.VarType.READER:
+                root_block._remove_var(var.name())
+
+        # change all `is_test` attributes to True
         for i in range(res.desc.num_blocks()):
             block = res.desc.block(i)
             for j in range(block.op_size()):
diff --git a/python/paddle/fluid/layers/io.py b/python/paddle/fluid/layers/io.py
index 77c61e086277444ea5f36c14e7ab8141052143fb..327ae309816344a0bcebfe70ffb59a00eab1d86f 100644
--- a/python/paddle/fluid/layers/io.py
+++ b/python/paddle/fluid/layers/io.py
@@ -443,9 +443,6 @@ def random_data_generator(low, high, shapes, lod_levels, for_parallel=True):
     main_prog_var = _copy_reader_var_(default_main_program().current_block(),
                                       startup_var)
 
-    if for_parallel:
-        main_prog_var = parallel(reader=main_prog_var)
-
     return monkey_patch_reader_methods(main_prog_var)