t_ids)

        >>> logits_next_token = outputs.logits  # logits to predict next token as usual
        >>> logits_ngram_next_tokens = outputs.logits_ngram  # logits to predict 2nd, 3rd, ... next tokens
        ```N)r„