Function clean_tree_text

seq2seq/constrained_seq2seq.py:38–39

(x)

def decode_tree_str(sequences: Union[List[int], List[List[int]], "np.ndarray", "torch.Tensor"],
                    tokenizer: PreTrainedTokenizer) -> List[str]:
    def clean_tree_text(x):
        return x.replace('<pad>', '').replace('<s>', '').replace('</s>', '').strip()

    sequences = np.where(sequences != -100, sequences, tokenizer.pad_token_id)