用OFA进行ocr训练任务

import os
import pandas as pd
import chardet
from PIL import Image
from datasets import Dataset
import tempfile
from modelscope.msdatasets import MsDataset
from modelscope.metainfo import Trainers
from modelscope.trainers import build_trainer
from modelscope.utils.constant import DownloadMode

with open('./ocr_labels_modelscope.csv', 'rb') as f:
result = chardet.detect(f.read())
data = pd.read_csv('./ocr_labels_modelscope.csv', encoding=result['encoding'])
ds = Dataset.from_pandas(data)
ds = MsDataset(ds)
print(next(iter(ds)))
文件格式：
image_id,text,image
000000000,硖,/mnt/workspace/images/000000000.jpg
000000001,饰,/mnt/workspace/images/000000001.jpg
000000002,晟,/mnt/workspace/images/000000002.jpg

同问，请问在文本行识别中如何使用MsDataset.load语句将图片和对应的标签传递到train_dataset中呢？