diff --git a/mindspore/ccsrc/transform/graph_ir/convert.cc b/mindspore/ccsrc/transform/graph_ir/convert.cc
index b1d84c33029..ffd80e76665 100644
--- a/mindspore/ccsrc/transform/graph_ir/convert.cc
+++ b/mindspore/ccsrc/transform/graph_ir/convert.cc
@@ -986,7 +986,7 @@ void DfGraphConvertor::BuildWhileSubGraph() {
     graph_name += "_body_graph";
     graph_out = GetWhileBodyOutputs();
   }
-  if (error_ == 0) {
+  if (error_ == SUCCESS) {
     if (df_graph_->GetName() != graph_name) {
       MS_LOG(DEBUG) << "convert anf graph name : " << df_graph_->GetName() << " to df graph name: " << graph_name;
     }
diff --git a/mindspore/ccsrc/transform/graph_ir/transform_util.cc b/mindspore/ccsrc/transform/graph_ir/transform_util.cc
index f0e62e5c36e..51e90f8d7d6 100644
--- a/mindspore/ccsrc/transform/graph_ir/transform_util.cc
+++ b/mindspore/ccsrc/transform/graph_ir/transform_util.cc
@@ -206,7 +206,7 @@ GeTensorPtr ConvertStringTensor(const MeTensorPtr &tensor, const std::string &fo
     char *string_element = new char[string_max_length];
     size_t string_length = 0;
     for (size_t i = 0; i < elements_num; i++) {
-      std::fill_n(string_element, string_max_length, '\0');
+      (void)std::fill_n(string_element, string_max_length, '\0');
       for (size_t j = 0; j < string_max_length; j++) {
         char char_element = data_ptr[i * string_max_length * single_char_offset + single_char_offset * j];
         if (static_cast<int>(char_element) == 0) {
diff --git a/mindspore/python/mindspore/common/api.py b/mindspore/python/mindspore/common/api.py
index 30414fdb4c4..0d869bbf217 100644
--- a/mindspore/python/mindspore/common/api.py
+++ b/mindspore/python/mindspore/common/api.py
@@ -468,6 +468,10 @@ def ms_function(fn=None, input_signature=None, hash_args=None, jit_config=None):
 
     This allows the MindSpore runtime to apply optimizations based on graph.
 
+    Note:
+        If `input_signature` is specified, each input of `fn` must be a Tensor. And the input arguments for `fn`
+        will not accept `**kwargs`.
+
     Args:
         fn (Function): The Python function that will be run as a graph. Default: None.
         input_signature (Tensor): The Tensor which describes the input arguments. The shape and dtype of the Tensor
diff --git a/mindspore/python/mindspore/train/model.py b/mindspore/python/mindspore/train/model.py
index 31212ff7469..657a5077bd4 100644
--- a/mindspore/python/mindspore/train/model.py
+++ b/mindspore/python/mindspore/train/model.py
@@ -713,7 +713,7 @@ class Model:
 
             dataset_helper.continue_send()
 
-            self._eval_durning_train(valid_infos, cb_params, list_callback)
+            self._eval_during_train(valid_infos, cb_params, list_callback)
 
             # In disaster recovery scenarios, need not to execute callbacks if this epoch executes failed.
             # Embedding cache server need not do epoch end callback, this process only run one step.
@@ -742,8 +742,8 @@ class Model:
 
         list_callback.on_train_end(run_context)
 
-    def _eval_durning_train(self, valid_infos, cb_params, list_callback):
-        """Exec eval durnning train process."""
+    def _eval_during_train(self, valid_infos, cb_params, list_callback):
+        """Exec eval during train process."""
         valid_dataset, valid_frequency, valid_dataset_sink_mode = valid_infos
         if valid_dataset and self._should_eval(cb_params.cur_epoch_num, valid_frequency):
             train_cur_step_num = cb_params.cur_step_num
@@ -921,7 +921,7 @@ class Model:
                 if should_stop:
                     break
 
-            self._eval_durning_train(valid_infos, cb_params, list_callback)
+            self._eval_during_train(valid_infos, cb_params, list_callback)
 
             train_dataset.reset()