17 lines
467 B
Python
17 lines
467 B
Python
from datasets import load_dataset
|
|
from modules.datapipeline import DataPipeline
|
|
|
|
if __name__ == "__main__":
|
|
chunk_size = 1000000
|
|
|
|
dataset = load_dataset(
|
|
"opencsg/chinese-cosmopedia",
|
|
data_files={"train": [f"data/000{i:02d}.parquet" for i in range(25)]}
|
|
)
|
|
|
|
pipeline = DataPipeline()
|
|
pipeline.process_dataset(
|
|
dataset_dict=dataset,
|
|
output_subdir="chinese-wiki-pretrain",
|
|
chunk_size=chunk_size,
|
|
) |