---
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)