update

app.py

@@ -89,7 +89,7 @@ def init_asset_plot(df):
         autosize=False,
         width=1000,
         height=700,
-        title=f"Accuracy per Asset"
+        title=f"Top 3 accuracies breakdown"
     )
     return fig
 

src/leaderboard/read_evals.py

@@ -73,10 +73,12 @@ 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]):
             if accs.size == 0 or any([acc is None for acc in accs]):
                 continue
             print(accs)
-            mean_acc = np.mean(accs) * 100.0
+            mean_acc = np.nanmean(accs) * 100.0
+            # mean_acc = np.mean(accs) * 100.0
             results[task.benchmark] = mean_acc
 
         return self(