how to use train_caption.json and train.txt ?

[Vincento-Wang]

is there any step missed of data preprocessing ?

I see the code, but caption is in the json file and it is a dict type, how to do like lines[2] ?

        with open(f'{self.meta_path}/{self.metadata[index]}', 'r') as f:
            lines = f.readlines()
        caption = lines[2].strip()
        lines = lines[3:]