Update masked_language_modeling.md (#24560)

See https://github.com/huggingface/transformers/issues/24546

docs/source/en/tasks/masked_language_modeling.md

@@ -168,7 +168,6 @@ You can now use a second preprocessing function to
 ...         k: [t[i : i + block_size] for i in range(0, total_length, block_size)]
 ...         for k, t in concatenated_examples.items()
 ...     }
-...     result["labels"] = result["input_ids"].copy()
 ...     return result
 ```