Methods of PLATO-KAG pre-training for other languages

[ak-sakamoto]

In README.md, the methods of fine-tuning is based on the Chinese and English pretrain models.
If we want to generate a pretrain model in other language, how should we train it?

https://github.com/PaddlePaddle/Knover/blob/develop/projects/PLATO-KAG/README.md