Questions about facebookresearch/XLM

I started experimenting with the XLM framework myself from https://github.com/facebookresearch/XLM.

I have the impression that this documentation is only for those who have implemented it, since there is no other documentation available (in case of an error you just have to unlock it yourself).

Nevertheless I have some questions. I have my own data (monolingual and paralingual) in the txt files, and I’d like to apply the decu preprocessing (fastBPE…) until I have the data understandable by the train.py script (BERT, XLM…).

At the same time, I’d like to understand all the formats (pth, …) used in the official website.

Each time I used the data from wikipedia and … wastes a lot of time