README.md

metadata

license: mit
language:
  - fa
  - en
library_name: adapter-transformers

I trained Llama2-7B after extending its tokenizer by 21,455 token on about 15B farsi text(common crawl, social, papers)