• Stars
    star
    123
  • Rank 290,145 (Top 6 %)
  • Language
    Jupyter Notebook
  • License
    Apache License 2.0
  • Created over 1 year ago
  • Updated over 1 year ago

Reviews

There are no reviews yet. Be the first to send feedback to the community and the maintainers!

Repository Details

A japanese finetuned instruction LLaMA

Japanese-Alpaca-LoRA

A Japanese finetuned instruction LLaMA.

In accordance with other models being referenced, this model is not available for commercial use and can only be used for research purposes.

References

This project is based on LLaMA, Stanford Alpaca, Alpaca LoRA, cabrita

Data

We translated the alpaca_data.json to japanese using ChatGPT API. We paid around US $45 to translate the full dataset to japanese. Translated data is available. (japanese_alpaca_data.json)

Finetuning

We just followed Alpaca LoRA and cabrita. We could run finetuning step using Google Colab PRO+. It took 6.5 hours for finetuning.

Example outputs

Good Examples

Input: 日本の首都はどこですか?
Alpaca-LoRA-Output: The capital of Japan is Tokyo.
Japanese-Alpaca-LoRA-Output: 「日本の首都は東京です。」
Input: 高齢の方が罹りやすい病気を3つ挙げてください。
Alpaca-LoRA-Output: Three diseases that are more common in the elderly are heart disease, stroke, and dementia.
Japanese-Alpaca-LoRA-Output:「高齢の方が罹りやすい病気には、心臓疾患、精神疾患、そして呼吸器疾患が含まれます。」

Bad Examples

Input: 大阪のオススメの食べ物を教えて
Alpaca-LoRA-Output: Osaka's favorite foods include takoyaki, okonomiyaki, kushikatsu, and yakitori.
Japanese-Alpaca-LoRA-Output:「大阪のオススメの食べ物には、オーストラリアの牛肉フルーツバター、オーストラリアの牛肉フルーツバター、オーストラリアの牛肉フルーツバター、
Input: 「3x + 2= 20」という方程式を解くとxはいくつになりますか
Alpaca-LoRA-Output:  x = 6
Japanese-Alpaca-LoRA-Output:「3x + 2 = 20」という方程式を解くと、xは2です。」