modift note for loss scale

mindspore/train/loss_scale_manager.py

@@ -36,7 +36,7 @@ class LossScaleManager:
 
 class FixedLossScaleManager(LossScaleManager):
     """
-    Fixed loss-scale manager.
+    Loss scale with a fixed value, inherits from LossScaleManager.
 
     Args:
         loss_scale (float): Loss scale. Note that if `drop_overflow_update` is set to False, the value of `loss_scale`
@@ -107,7 +107,7 @@ class FixedLossScaleManager(LossScaleManager):
 
 class DynamicLossScaleManager(LossScaleManager):
     """
-    Dynamic loss-scale manager.
+    Loss scale that dynamically adjusts itself, inherits from LossScaleManager.
 
     Args:
         init_loss_scale (float): Initialize loss scale. Default: 2**24.

model_zoo/official/nlp/bert/run_eval_onnx.py

@@ -56,9 +56,14 @@ def do_eval_onnx(dataset=None, num_class=15, assessment_method="accuracy"):
         raise ValueError("Assessment method not supported, support: [accuracy, f1, mcc, spearman_correlation]")
 
     columns_list = ["input_ids", "input_mask", "segment_ids", "label_ids"]
-    onnx_file_name = args_opt.export_file_name + '.onnx'
+    onnx_file_name = args_opt.export_file_name
+    if not args_opt.export_file_name.endswith('.onnx'):
+        onnx_file_name = onnx_file_name + '.onnx'
     if not os.path.isabs(onnx_file_name):
         onnx_file_name = os.getcwd() + '/' + onnx_file_name
+    if not os.path.exists(onnx_file_name):
+        raise ValueError("ONNX file not exists, please check onnx file has been saved and whether the "
+                         "export_file_name is correct.")
     sess = rt.InferenceSession(onnx_file_name)
     input_name_0 = sess.get_inputs()[0].name
     input_name_1 = sess.get_inputs()[1].name
@@ -88,8 +93,6 @@ def run_classifier_onnx():
     """run classifier task for onnx model"""
     if args_opt.eval_data_file_path == "":
         raise ValueError("'eval_data_file_path' must be set when do onnx evaluation task")
-    if args_opt.onnx_file_path == "":
-        raise ValueError("'onnx_file_path' must be set when do onnx evaluation task")
     assessment_method = args_opt.assessment_method.lower()
     ds = create_classification_dataset(batch_size=args_opt.eval_batch_size, repeat_count=1,
                                        assessment_method=assessment_method,