No public description

PiperOrigin-RevId: 570043009

Author: T5 Team

t5/evaluation/eval_utils.py

@@ -233,7 +233,7 @@ def metric_group_max(df, metric_names=None):
   for group, metrics in group_to_metrics.items():
     if not all(m in df for m in metrics):
       continue
-    group_df[group] = df[metrics].mean(axis=1)
+    group_df[group] = df[list(metrics)].mean(axis=1)
   # Need to replace nan with large negative value for idxmax
   group_max_step = group_df.fillna(-1e9).idxmax(axis=0)
   metric_max = pd.Series()