feat(dataset): 更新数据集处理和保存逻辑

This commit is contained in:
2025-07-20 11:21:18 +08:00
parent 365086f102
commit fcb5c2bcc8
3 changed files with 17 additions and 2 deletions
+15
View File
@@ -0,0 +1,15 @@
from utils import dump_pkl_files, fetch_files
from tokenizer import BpeTokenizer
import os
if __name__ == "__main__":
tokenizer = BpeTokenizer("tokenizer.json")
base_dir = [
os.path.join("dataset", "belle-sft"),
]
base_out_dir = "pkl_output"
files = []
for dir_path in base_dir:
files.extend(fetch_files(dir_path))
dump_pkl_files(tokenizer, files, base_out_dir, packing_size=2048)