F

src/leaderboard/read_evals.py

@@ -118,7 +118,8 @@ class EvalResult:
 
     def to_dict(self, tasks):
         """Converts the Eval Result to a dict compatible with our dataframe display"""
-        average = sum([v for v in self.results.values() if v is not None]) / len(tasks)
+        avg_metrics = [v for k, v in self.results.items() if v is not None and " " not in k]
+        average = sum(avg_metrics) / len(avg_metrics)
         data_dict = {
             "eval_name": self.eval_name,  # not a column, just a save name,
             AutoEvalColumn.architecture.name: self.architecture,