from datasets import load_dataset
import json
import random

dataset = load_dataset("ethos", "binary")
id2label = dataset['train'].features['label'].names

rows = [{'text': row['text'], 'label': row['label'], 'label_text': id2label[row['label']].replace("_", " ")} for row in dataset['train']]

random.seed(42)
random.shuffle(rows)

num_test = 400

datasplits = {'train': rows[num_test:],  'test': rows[0:num_test]}


for split in ['train', 'test']:
    with open(f'{split}.jsonl', 'w') as fOut:
        for row in datasplits[split]:
            fOut.write(json.dumps(row)+"\n")