Bugfix: 0 is now considered a valid step number in log_time_series_metrics

simonhessner · simonhessner · commit f6e8a6a4e20f · 2024-01-12T21:36:02.000Z
Previously, when log_time_series_metrics is called without a step value it defaults to None. The function internally checked `if not step` and
in this case, determined a step number. The problem is that 0 is falsy and therefore the function turns step=0 into step=1 which is a problem if
the function is later called with step=1 (in that case the first logged values are overwritten)

Checking for `if step is None` fixes the problem.
diff --git a/google/cloud/aiplatform/metadata/experiment_run_resource.py b/google/cloud/aiplatform/metadata/experiment_run_resource.py
@@ -961,7 +961,7 @@ def log_time_series_metrics(
 
         self._soft_create_time_series(metric_keys=set(metrics.keys()))
 
-        if not step:
+        if step is None:
             step = self._largest_step or self._get_latest_time_series_step()
             step += 1
             self._largest_step = step
