fix the bug of step_trace incorrect content in inference scene

mindinsight/profiler/analyser/step_trace_analyser.py

@@ -217,13 +217,24 @@ class StepTraceAnalyser(BaseAnalyser):
         start_point = row_info_dict.get('start_point', 0)
         fp_point = row_info_dict.get('fp_point', 0)
         bp_point = row_info_dict.get('bp_point', 0)
-        points = [
+        points_part = [
             self._construct_time_point(
                 'iteration_interval', 0, row_info_dict.get('iteration_interval', 0)),
-            self._construct_time_point(
-                'fp_and_bp', fp_point - start_point, row_info_dict.get('fp_and_bp', 0)),
-            self._construct_time_point('tail', bp_point - start_point, row_info_dict.get('tail', 0))
         ]
+        # if fp key exist, inference scene
+        if 'fp' in row_info_dict.keys():
+            points = [
+                self._construct_time_point(
+                    'fp', fp_point - start_point, row_info_dict.get('fp', 0)),
+            ]
+        # training scene
+        else:
+            points = [
+                self._construct_time_point(
+                    'fp_and_bp', fp_point - start_point, row_info_dict.get('fp_and_bp', 0)),
+                self._construct_time_point('tail', bp_point - start_point, row_info_dict.get('tail', 0))
+            ]
+        points = points_part + points
         return points
 
     def _get_reduce_time_in_order(self, row_info_dict):

mindinsight/profiler/common/util.py

@@ -125,17 +125,29 @@ def get_summary_for_step_trace(average_info, header):
     total_time = get_field_value(average_info, 'total', header)
     iteration_interval = get_field_value(average_info, 'iteration_interval',
                                          header)
-    fp_and_bp = get_field_value(average_info, 'fp_and_bp', header)
-    tail = get_field_value(average_info, 'tail', header)
-    summary = {
+    summary_part = {
         'total_time': total_time,
         'iteration_interval': iteration_interval,
         'iteration_interval_percent': calculate_percent(iteration_interval, total_time),
-        'fp_and_bp': fp_and_bp,
-        'fp_and_bp_percent': calculate_percent(fp_and_bp, total_time),
-        'tail': tail,
-        'tail_percent': calculate_percent(tail, total_time)
     }
+    # training scene data for ui display
+    if 'fp_and_bp' in header:
+        fp_and_bp = get_field_value(average_info, 'fp_and_bp', header)
+        tail = get_field_value(average_info, 'tail', header)
+        summary = {
+            'fp_and_bp': fp_and_bp,
+            'fp_and_bp_percent': calculate_percent(fp_and_bp, total_time),
+            'tail': tail,
+            'tail_percent': calculate_percent(tail, total_time)
+        }
+    # inference scene data for ui display
+    else:
+        fp = get_field_value(average_info, 'fp', header)
+        summary = {
+            'fp': fp,
+            'fp_percent': calculate_percent(fp, total_time)
+        }
+    summary.update(summary_part)
     return summary
 
 

mindinsight/profiler/common/validator/validate.py

@@ -250,12 +250,12 @@ def validate_ui_proc(proc_name):
 
     Args:
         proc_name (str): The proc name to query. Acceptable value is in
-        [`iteration_interval`, `fp_and_bp`, `tail`].
+        [`iteration_interval`, `fp_and_bp`, `fp`, `tail`].
 
     Raises:
         ProfilerParamValueErrorException: If the proc_name is invalid.
     """
-    accept_names = ['iteration_interval', 'fp_and_bp', 'tail']
+    accept_names = ['iteration_interval', 'fp_and_bp', 'fp', 'tail']
     if proc_name not in accept_names:
         log.error("Invalid proc_name. The proc_name for restful api is in %s", accept_names)
         raise ProfilerParamValueErrorException(f'proc_name should be in {accept_names}.')