LLMclass / README.md
yuwei1289's picture
Create README.md
c5adbc3 verified
|
Raw
History Blame Contribute Delete
1.2 kB
import os
import json
import pandas as pd
folder_path = "双语数据\史记\七十列传"
def get_files(folder_path):
subfolders = [f.path for f in os.scandir(folder_path) if f.is_dir()]
print(subfolders)
dataset = []
source_file = "source.txt"
target_file = "target.txt"
for x in subfolders:
with open(os.path.join(x,source_file) , "r", encoding="utf-8") as f:
source_content = f.read()
with open(os.path.join(x,target_file), "r", encoding="utf-8") as f:
target_content = f.read()
source_content = source_content.split("\n")
target_content = target_content.split("\n")
for i in range(len(source_content)):
dataset.append([source_content[i], target_content[i]])
return dataset
dataset = get_files(folder_path)
# add one column "instruction" with the content "请把古文翻译成现代汉语" to the dataset
df = pd.DataFrame(dataset, columns=["source", "target"])
df["instruction"] = "请把现代汉语翻译成古文"
df.rename(columns={"source": "output", "target": "input"}, inplace=True)
print(len(df))
df.to_json("dataset.jsonl", orient="records", lines=True, force_ascii=False)