Swallow - work4ai

Swallow

https://gyazo.com/690e2e63942a71c33daffd8b88f9d60c

https://tokyotech-llm.github.io/swallow-llamaSwallow

https://huggingface.co/tokyotech-llm/Swallow-7b-hfTokyoTech-LLM/Swallow-7b-hf

https://huggingface.co/tokyotech-llm/Swallow-7b-instruct-hfTokyoTech-LLM/Swallow-7b-instruct-hf

https://huggingface.co/tokyotech-llm/Swallow-13b-hfTokyoTech-LLM/Swallow-13b-hf

https://huggingface.co/tokyotech-llm/Swallow-13b-instruct-hfTokyoTech-LLM/Swallow-13b-instruct-hf

https://huggingface.co/tokyotech-llm/Swallow-70b-hfTokyoTech-LLM/Swallow-70b-hf

https://huggingface.co/tokyotech-llm/Swallow-70b-instruct-hfTokyoTech-LLM/Swallow-70b-instruct-hf

LLama 2に日本語の語彙を追加して事前学習をした

ベンチマーク

https://tokyotech-llm.github.io/images/7B_ja.svghttps://tokyotech-llm.github.io/images/13B_ja.svg

https://tokyotech-llm.github.io/images/70B_ja.svghttps://tokyotech-llm.github.io/images/swallow_ja.svg

比較：

Japanese Stable LM Beta

東京工業大学

TSUBAMEじゃないんかいwogikaze.icon

#大規模言語モデル構築支援プログラム

LLAMA 2 Community License

→Llama-3-Swallow