The models are based on Grammarly's Yahoo Answers Formality Corpus (GYAFC)
Description of this corpus can be found in the paper below:
Dear Sir or Madam, May I introduce the GYAFC Corpus: Corpus, Benchmarks and Metrics for Formality Style Transfer
Sudha Rao (raosudha@cs.umd.edu) and Joel Tetreault (tetreaul@gmail.com)
To appear in the proceedings of The 2018 North American Association of Computational Lingusitics (NAACL 2018)

The GYAFC corpus was created using the Yahoo Answers corpus: L6 - Yahoo! Answers Comprehensive Questions and Answers version 1.0 . This Yahoo Answers corpus can be requested free of charge for research purposes.
Access to our GYAFC dataset will require users to first gain access to this Yahoo Answers corpus.

Once you have gained access to the L6 corpus, please forward the acknowledgment to Joel Tetreault (tetreaul@gmail.com), along with your affiliation and a short description of how you will be using the data, and we will provide access to our GYAFC corpus.

Once you have the acknowledgment of Joel Tetreault forward the email to Cristhiam Tirado (u20151c149@upc.edu.pe) to receive access to the weights.

Please let us know if you have any questions.