Update app.py and evaluate.py with latest changes

app.py

@@ -120,7 +120,7 @@ with gr.Blocks(title="IndoNLP Dashboard") as demo:
                     new_entry = {"task": "Sentiment", "model": model_key, **res['metrics']}
                     # append and save
                     global leaderboard
-                    leaderboard = leaderboard.append(new_entry, ignore_index=True)
+                    leaderboard = pd.concat([leaderboard, pd.DataFrame([new_entry])], ignore_index=True)
                     save_leaderboard(leaderboard, LEADERBOARD_PATH)
                     # build confusion matrix plot
                     cm = res.get('confusion_matrix')

metrics/evaluate.py

@@ -158,7 +158,7 @@ def evaluate_qa(file, model_key: str) -> Dict[str, Any]:
 
 
 def load_leaderboard(path: str):
-    if not os.path.exists(path):
+    if not os.path.exists(path) or os.path.getsize(path) == 0:
         # default
         df = pd.DataFrame([{"task": "Sentiment", "model": "indobert", "f1": 0.88, "accuracy": 0.87}])
         df.to_json(path, orient='records')
@@ -167,4 +167,4 @@ def load_leaderboard(path: str):
 
 
 def save_leaderboard(df, path: str):
-    df.to_json(path, orient='records')
+    return pd.read_json(path)