Set power_measurement_stop to last measurement

mlcommons · May 21, 2024 · 4054425 · 4054425
1 parent ced0239
commit 4054425
Showing 1 changed file with 3 additions and 0 deletions.
diff --git a/mlperf_logging/result_summarizer/result_summarizer.py b/mlperf_logging/result_summarizer/result_summarizer.py
@@ -516,6 +516,7 @@ def _compute_total_power(benchmark_folder, result_file, time_to_train, ruleset):
 
 def _compute_power_node(loglines, time_to_train):
     prev_timestamp = 0
+    max_timestamp = 0
     power_start = 0
     power_stop = 0
     agg_power = 0
@@ -527,13 +528,15 @@ def _compute_power_node(loglines, time_to_train):
         if logline.key == "power_reading":
             agg_power += (logline.value['value'] * (logline.timestamp - prev_timestamp))
             prev_timestamp = logline.timestamp
+            max_timestamp = max(max_timestamp, logline.timestamp)
         if logline.key == "power_measurement_stop":
             power_stop = logline.timestamp
             break
         if logline.key == "conversion_eff":
             conversion_eff = logline.value['value']
 
     # Compute the result, convert ms to s
+    power_stop = max(power_stop, max_timestamp)
     result = conversion_eff * agg_power * time_to_train / (power_stop - power_start) / 1000
     return result