Datasets:

Modalities:
Text
Formats:
json
Languages:
Japanese
ArXiv:
Libraries:
Datasets
pandas
License:
masanorihirano's picture
Update README.md
ffc2d16
|
raw
history blame
2.18 kB
metadata
license: cc-by-sa-4.0
language:
  - ja
size_categories:
  - 1M<n<10M

llm-japanese-dataset

LLM構築用の日本語チャットデータセット

主に,英語で構築されたLLMモデルなどに対して,チャット(Instruction)応答タスクに関してLoRAなどでチューニングするために使用できます.

※様々な公開言語資源を利用させていただきました.関係各位にはこの場を借りて御礼申し上げます.

データの詳細

データの詳細は,以下の論文を参照してください.

なお,Citationには,よろしければ,以下をご利用ください.

@preprint{Hirano2023-llmj,
  title={{llm-japanese-dataset v0: Construction of Japanese Chat Dataset for Large Language Models and its Methodology}},
  autor={Masanori HIRANO and Masahiro SUZUKI and Hiroki SAKAJI},
  doi={10.48550/arXiv.2305.12720},
  archivePrefix={arXiv},
  arxivId={2305.12720},
  year={2023}
}

共同研究,データ提供,各種支援,その他問い合わせは,izumi-llm@socsim.org へ.

How to use

from datasets import load_dataset
dataset = load_dataset("izumi-lab/llm-japanese-dataset", revision="main")
dataset = load_dataset("izumi-lab/llm-japanese-dataset", revision="a.b.c") # for specific version

version 0.1.0 contains bugs version 0.1.1 contains 8,393,726 data (bug fixed)

For more details, see: https://github.com/masanorihirano/llm-japanese-dataset

LICENSE

CC-BY-SA 4.0 (For more details, see: LICENSE, NOTICE.md, NOTICE2.md)

Note

MIT License version is also available on the github release page https://github.com/masanorihirano/llm-japanese-dataset/releases

To see more latest information, please go to llm.msuzuki.me.