| 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) | |
| ``` |