This dataset is sampled from `wikitext/wikitext-2-v1/train`. | |
Codes to generate this dataset: | |
```python | |
import datasets | |
dataset = datasets.load_dataset('wikitext', 'wikitext-2-v1') | |
selected = [] | |
i = -1 | |
while len(selected) < 24: | |
i += 1 | |
text = dataset['train'][i]['text'] | |
if 8 < len(text.split(' ')) <= 16 and '=' not in text: | |
selected.append(i) | |
tiny_dataset = dataset['train'].select(selected) | |
``` |