README.md

metadata

language:
  - ko
tags:
  - llama-2
  - instruct
  - instruction
pipeline_tag: text-generation
license: llama2
datasets:
  - beomi/KoAlpaca-v1.1a
  - kyujinpy/KOpen-platypus

Llama-2-ko-instruct-13B

Model Details

• Base Model: LLaMA-2-koen-13B

Datasets

• Added some English to Korean translation data based on the KOpen-platypus and KoAlpaca datasets. Translations utilized AWS blog content that I translated myself.
• Extracted only sentences longer than 100 characters and removed similar sentences with KoSimCSE (daekeun-ml/KoSimCSE-supervised-kobigbird-roberta-large)
• Created category-specific prompts that encourage AI to answer despite hallucination for future RLHF (Reinforcement Learning From Human Feedback) or DPO (Direct Preference Optimization) tuning.

License

• Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International Public License, under LLAMA 2 COMMUNITY LICENSE AGREEMENT

This model was created as a personal experiment, unrelated to the organization I work for.