Fix MASK token noising

app.py

@@ -115,7 +115,7 @@ def noisify_answer(input_ids, answer_start, threshold=1.0, eot_weight=1.0, clust
     noise = rng.choice(np.arange(vocab_size), size=len(noised_indices), p=mixed_probs)
     for idx, val in zip(noised_indices, noise):
         noised[idx] = val
-        noised[idx] = tokenizer.encode('MASK', add_special_tokens = False)
+        noised[idx] = tokenizer.encode('MASK', add_special_tokens = False)[0]
 
     return noised, noised_indices