fcd6b2bc97933cb686977e8b95e2bef2e99b5ea9,run_tgen.py,,eval_tokens,#Any#Any#,533

Before Change

def eval_tokens(eval_tokens, gen_tokens):
    // TODO TODO TODO
    raise NotImplementedError()


def rerank_cl_eval(args):

After Change

    log_info("BLEU score: %.4f" % (evaluator.bleu() * 100))

    evaluator = Evaluator()
    for pred_sent, gold_sents in zip(gen_tokens, eval_tokens):
        for gold_sent in gold_sents:  // effectively an average over all gold paraphrases
            evaluator.append(gold_sent, pred_sent)

    log_info("TOKEN precision: %.4f, Recall: %.4f, F1: %.4f" % evaler.p_r_f1(EvalTypes.TOKEN))
    log_info("Sentence length stats:\n * GOLD %s\n * PRED %s\n * DIFF %s" % evaler.size_stats())
    log_info("Common subphrase stats:\n -- SIZE: %s\n -- ΔGLD: %s\n -- ΔPRD: %s" %
             evaler.common_substruct_stats())

Instances