wangbase.py 文件源码

def _devset_cv(self, a_y_train, a_n_dev, a_n_folds):
        """Generate train-test split from training and development data.

        Args:
          a_y_train (list[int]):
            list of training instances' tags
          a_n_dev (int):
            number of devset instances
          a_n_folds (int):
            number of folds

        Returrns:
          list[tuple]: list of training/testing folds

        """
        folds = []
        n_train = len(a_y_train)
        dev_ids = [n_train + i for i in xrange(a_n_dev)]
        # create stratified K-folds over the training data
        skf = StratifiedKFold(a_y_train, a_n_folds)
        for train_ids, test_ids in skf:
            folds.append((train_ids,
                          np.concatenate((test_ids, dev_ids))))
        return folds