| 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") |