Function compute_metrics

run_seq2seq.py:612–654

(eval_preds)

        return preds, labels

    def compute_metrics(eval_preds):
        preds, labels = eval_preds
        if isinstance(preds, tuple):
            preds = preds[0]
        decoded_preds = tokenizer.batch_decode(
            preds, skip_special_tokens=False)
        if data_args.ignore_pad_token_for_loss:
            # Replace -100 in the labels as we can't decode them.
            labels = np.where(labels != -100, labels, tokenizer.pad_token_id)
        decoded_labels = tokenizer.batch_decode(
            labels, skip_special_tokens=False)

        def clean_str(x_str):
            for to_remove_token in to_remove_token_list:
                x_str = x_str.replace(to_remove_token, '')
            return x_str.strip()

        decoded_preds = [clean_str(x) for x in decoded_preds]
        decoded_labels = [clean_str(x) for x in decoded_labels]

        # Some simple post-processing
        # decoded_preds, decoded_labels = postprocess_text(decoded_preds, decoded_labels)

        # if metric_name == "rouge":
        #     result = metric.compute(predictions=decoded_preds, references=decoded_labels, use_stemmer=True)
        #     # Extract a few results from ROUGE
        #     result = {key: value.mid.fmeasure * 100 for key, value in result.items()}
        # else:
        #     result = metric.compute(predictions=decoded_preds, references=decoded_labels)
        #     result = {"bleu": result["score"]}

        result = get_extract_metrics(
            pred_lns=decoded_preds,
            tgt_lns=decoded_labels,
            label_constraint=decoding_type_schema,
            decoding_format=data_args.decoding_format,
        )

        prediction_lens = [np.count_nonzero(
            pred != tokenizer.pad_token_id) for pred in preds]
        result["gen_len"] = np.mean(prediction_lens)
        result = {k: round(v, 4) for k, v in result.items()}
        return result

    # Initialize our Trainer
    trainer = ConstraintSeq2SeqTrainer(