Fix: handle numpy labels correctly for small HF demo datasets

training_utils.py

@@ -66,7 +66,8 @@ def train_model_cv(df, features, target='Label', n_splits=5, n_estimators=100, m
     all_val_labels = []
 
     X_arr = X.values
-    y_arr = y.values
+    y_arr = np.asarray(y)
+
 
     for fold, (train_idx, val_idx) in enumerate(skf.split(X_arr, y_arr), start=1):
         clf = RandomForestClassifier(