Adapting Llama-2 to Estonian

This repository contains the fine-tuning, inference and data formating scripts for fine-tuning and continued-pretraining of Llama-2 for Estonian.

The scripts directory contains example scripts for:

• scripts/training for example training scripts.
• scripts/chat_data for formating instructions into Estonian/English chat format.
• scripts/translation_data for bitext creation and formating into Estonian/English instructions.

For instructions used to train the model:

• Alpaca-est (ours)
• Alpaca-cleaned
• open-instruct

Trained model checkpoints:

• Llammas (conversational/instruction-tuned)
• Llammas-base
• Llammas-translate (conversational/instruction-tuned with focus on translation)

@misc{kuulmets2024teaching,
      title={Teaching Llama a New Language Through Cross-Lingual Knowledge Transfer}, 
      author={Hele-Andra Kuulmets and Taido Purason and Agnes Luhtaru and Mark Fishel},
      year={2024},
      eprint={2404.04042},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}