| 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") | |