update

src/leaderboard/read_evals.py

@@ -74,7 +74,7 @@ class EvalResult:
             # We average all scores of a given metric (not all metrics are present in all files)
             accs = np.array([v.get(task.metric, None) for k, v in data["results"].items() if task.benchmark == k])
             if accs.size == 0 or any([acc is None for acc in accs]):
-                mean_acc = 0.0
+                mean_acc = None
             else:
                 mean_acc = np.nanmean(accs) * 100.0
             # if accs.size == 0:
@@ -113,7 +113,9 @@ class EvalResult:
 
     def to_dict(self, task_class):
         """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(task_class)
+        scores = [v for v in self.results.values() if v is not None]
+        average = sum(scores) / len(scores)
+        # average = sum([v for v in self.results.values() if v is not None]) / len(task_class)
         data_dict = {
             "eval_name": self.eval_name,  # not a column, just a save name,
             AutoEvalColumn.precision.name: self.precision.value.name,