from datasets import load_dataset
import json
import random

dataset = load_dataset("SetFit/amazon_counterfactual", "en-ext")

examples = list(dataset['train'])+list(dataset['validation'])+list(dataset['test'])
random.seed(42)
random.shuffle(examples)


num_test = 5000
data_splits = {'test': examples[0:num_test], 'train': examples[num_test:]}


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