train_with_summary.txt train_tokenized.txt

[world4jason]

有沒有這兩個的連結?
參考一下格式
謝謝

[Clay]

Maybe you can refer： https://zhuanlan.zhihu.com/p/113869509

The data format is similar to the following example:
{"summarization": "xxxxxxxxx", "article": "aaaaaaaaa"}

You can use json.dumps() to convert data to string data type and save it, using '\n' to split data. (Because the source code is using json.loads() to load the training data)