refactor: 重命名pretrain 和 sft 脚本

This commit is contained in:
2025-07-21 19:58:15 +08:00
parent c2821c150a
commit f9ce67e03c
3 changed files with 1 additions and 1 deletions
+21
View File
@@ -0,0 +1,21 @@
from utils import dump_pkl_files, fetch_files
from tokenizer import BpeTokenizer
import os
def processor(intput_str: str):
return f"{intput_str} <eos>"
if __name__ == "__main__":
tokenizer = BpeTokenizer("tokenizer.json")
base_dir = [
os.path.join("dataset", "chinese-c4"),
os.path.join("dataset", "english-fineweb"),
os.path.join("dataset", "english-wiki"),
os.path.join("dataset", "chinese-wiki"),
]
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, processor)