LLaMa2 - 7B Chat models, extend vocab size to 44800 for Vietnamese understanding.
Continual Pre-Train with 2B Vietnames Tokens aligned from VnNews Corpus, 10K vnthuquan books, wikipedia_vi
Fine-Tuning with infCapital/viet-llama2-ft-tiny dataset, the combination of vaious dataset then translated into Vietnamese using OpenAI GPT-3
For more information: email me at duyhunghd6@gmail.com | http://fb.com/hungbui2013
- Downloads last month
- 1,095