| from datasets import load_dataset | |
| import os | |
| from kowiki import process_text | |
| for date in ['20230420', '20230501', '20230520', '20230601', '20230620', '20230701', '20230720', '20230801']: | |
| dataset = load_dataset("psyche/kowiki", date, streaming=False, split="train") | |
| def process_function(example): | |
| example["text"] = process_text(example["text"]) | |
| return example | |
| dataset = dataset.map(process_function) | |
| os.makedirs(f"data/{date}", exist_ok=True) | |
| for i in range(10): | |
| dataset.shard(num_shards=10, index=i).to_parquet(f"data/{date}/train-{i:05d}-of-{10:05d}.parquet") | |