modified read_evals.py

src/leaderboard/read_evals.py

@@ -65,11 +65,8 @@ 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]):
-                continue
 
-            mean_acc = np.mean(accs) * 100.0
-            results[task.benchmark] = mean_acc
+            results[task.benchmark] = accs 
 
         return self(
             eval_name=result_key,