Llama 3.1 Swallow 8B Instruct V0.2

🚀 Llama 3.1 Swallow - Built with Llama

Llama 3.1 Swallow is a series of large language models that enhance Japanese language capabilities while retaining English proficiency. It's built through continual pre - training and instruction - tuning.

🚀 Quick Start

To start using the Llama 3.1 Swallow model, you first need to install the vllm library:

pip install vllm

Here is a basic Python code example to generate text using the model:

from transformers import AutoTokenizer
from vllm import LLM, SamplingParams

model_name = "tokyotech-llm/Llama-3.1-Swallow-8B-Instruct-v0.2"

tokenizer = AutoTokenizer.from_pretrained(model_name)
llm = LLM(
    model=model_name,
    tensor_parallel_size=1,
)

sampling_params = SamplingParams(
    temperature=0.6, top_p=0.9, max_tokens=512, stop="<|eot_id|>"
)


message = [
    {"role": "system", "content": "あなたは誠実で優秀な日本人のアシスタントです。"},
    {
        "role": "user",
        "content": "東京の紅葉した公園で、東京タワーと高層ビルを背景に、空を舞うツバメと草地に佇むラマが出会う温かな物語を書いてください。",
    },
]
prompt = tokenizer.apply_chat_template(
    message, tokenize=False, add_generation_prompt=True
)

output = llm.generate(prompt, sampling_params)

print(output[0].outputs[0].text)

✨ Features

• Multilingual Capability: Llama 3.1 Swallow supports both Japanese and English, with enhanced Japanese language capabilities compared to the original Llama 3.1.
• Continual Pre - training: The model is built by continual pre - training on the Meta Llama 3.1 models, using a large amount of Japanese and English data.
• Instruction - tuned Variants: Instruction - tuned models are available, which are built by supervised fine - tuning on synthetic data specially built for Japanese.

📦 Installation

Install the vllm library using the following command:

pip install vllm

💻 Usage Examples

Basic Usage

from transformers import AutoTokenizer
from vllm import LLM, SamplingParams

model_name = "tokyotech-llm/Llama-3.1-Swallow-8B-Instruct-v0.2"

tokenizer = AutoTokenizer.from_pretrained(model_name)
llm = LLM(
    model=model_name,
    tensor_parallel_size=1,
)

sampling_params = SamplingParams(
    temperature=0.6, top_p=0.9, max_tokens=512, stop="<|eot_id|>"
)


message = [
    {"role": "system", "content": "あなたは誠実で優秀な日本人のアシスタントです。"},
    {
        "role": "user",
        "content": "東京の紅葉した公園で、東京タワーと高層ビルを背景に、空を舞うツバメと草地に佇むラマが出会う温かな物語を書いてください。",
    },
]
prompt = tokenizer.apply_chat_template(
    message, tokenize=False, add_generation_prompt=True
)

output = llm.generate(prompt, sampling_params)

print(output[0].outputs[0].text)

📚 Documentation

Swallow Model Index

Model Details

Model Performance

Japanese tasks

English tasks

MT - Bench JA

Evaluation Benchmarks

Japanese evaluation benchmarks

We used llm - jp - eval(v1.3.0), JP Language Model Evaluation Harness(commit #9b42d41) and Code Generation LM Evaluation Harness(commit #0261c52). The details are as follows:

• Multiple - choice question answering (JCommonsenseQA [Kurihara et al., 2022])
• Open - ended question answering (JEMHopQA [Ishii et al., 2024])
• Open - ended question answering (NIILC [関根, 2003])
• Machine reading comprehension (JSQuAD [Kurihara et al., 2022])
• Automatic summarization (XL - Sum [Hasan et al., 2021])
• Machine translation (WMT2020 ja - en [Barrault et al., 2020])
• Machine translation (WMT2020 en - ja [Barrault et al., 2020])
• Mathematical reasoning (MGSM [Shi et al., 2023])
• Academic exams (JMMLU [尹ら, 2024])
• Code generation (JHumanEval [佐藤ら, 2024])

English evaluation benchmarks

We used the Language Model Evaluation Harness(v.0.4.2) and Code Generation LM Evaluation Harness(commit #0261c52). The details are as follows:

• Multiple - choice question answering (OpenBookQA [Mihaylov et al., 2018])
• Open - ended question answering (TriviaQA [Joshi et al., 2017])
• Machine reading comprehension (SQuAD2 [Rajpurkar et al., 2018])
• Commonsense reasoning (XWINO [Tikhonov and Ryabinin, 2021])
• Natural language inference (HellaSwag [Zellers et al., 2019])
• Mathematical reasoning (GSM8K [Cobbe et al., 2021])
• Reasoning (BBH (BIG - Bench - Hard) [Suzgun et al., 2023])
• Academic exams (MMLU [Hendrycks et al., 2021])
• Code generation (HumanEval [Chen et al., 2021])

MT - Bench JA

We used Japanese MT - Bench to assess the capabilities of multi - turn dialogue with the following settings:

• Implementation: FastChat [Zheng+, 2023] (commit #e86e70d0)
• Question: Nejumi LLM - Leaderboard NEO, mtbench_ja_question_v3
• Reference Answer: Nejumi LLM - Leaderboard NEO, mtbench_ja_referenceanswer_v1
• Prompt for Judge: Nejumi LLM - Leaderboard NEO, mtbench_ja_prompt_v1
• Judge: gpt - 4 - 1106 - preview
• Scoring: Absolute scale normalized to a 0 - 1 range, averaged over five runs.

Training Datasets

Instruction Tuning

The following datasets were used for the instruction tuning:

• Japanese
  • Llama-3.1-LMSYS-Chat-1M-Synth-Ja: Single - turn Japanese instruction dataset synthesized and derived from lmsys - chat - 1m [Zhang+, ICLR24]). First - turn user instructions were translated into Japanese via DeepL (machine translation), and assistant responses were generated using Llama - 3.1 - 405B - Instruct. Llama - 3.1 - 70B - Instruct served as a judge for rejection sampling (n = 6). Conversations containing personally identifiable information (PII) and template - based user instructions were removed. Duplicate instructions were removed.
  • Swallow - Magpie - Ultra - v0.1: A Japanese variant of the filtered - magpie - ultra - en dataset, translated into Japanese by gemma - 2 - 27b - it.
  • Swallow - Gemma - Magpie - v0.1: A Japanese synthetic instruction tuning dataset from scratch, generated by gemma - 2 - 27b - it. User instructions were created with prompts specific to each topic, and assistant responses were generated for these instructions. The conversations were then heuristically filtered for quality and length.
• English
  • Llama-3.1-LMSYS-Chat-1M-Synth-En: The creation process is similar to Llama - 3.1 - LMSYS - Chat - 1M - Synth - Ja, but this version uses the original English user instructions. The assistant responses were generated in English as well. Rejection sampling was not applied for this version.
  • filtered - magpie - ultra - en: A subset of the magpie - ultra dataset, developed following the MAGPIE recipe [Xu+, arXiv24] using Llama - 3.1 - 405B - Instruct. This subset includes only samples rated as 'average,' 'good,' or 'excellent.'

🔧 Technical Details

The Llama 3.1 Swallow models are built through continual pre - training on the Meta Llama 3.1 models. Approximately 200 billion tokens sampled from a large Japanese web corpus (Swallow Corpus Version 2), Japanese and English Wikipedia articles, and mathematical and coding contents are used for continual pre - training. The instruction - tuned models are built by supervised fine - tuning on synthetic data specially built for Japanese.

📄 License

META LLAMA 3.1 COMMUNITY LICENSE and Gemma Terms of Use

👥 Authors

Here are the team members:

• From Tokyo Institute of Technology Okazaki Laboratory:
  • Naoaki Okazaki
  • Sakae Mizuki
  • Youmi Ma
  • Koki Maeda
  • Kakeru Hattori
  • Masanari Ohi
  • Taihei Shiotani
  • Koshiro Saito
• From Tokyo Institute of Technology YOKOTA Laboratory:
  • Rio Yokota
  • Kazuki Fujii
  • Taishi Nakamura
  • Takumi Okamoto
  • Ishida Shigeki
• From Artificial Intelligence Research Center, AIST, Japan:
  • Hiroya Takamura

📖 How to cite

If you find our work helpful, please feel free to cite these papers:

@inproceedings{Fujii:COLM2024,
   title={Continual Pre-Training for Cross-Lingual LLM Adaptation:
Enhancing Japanese Language Capabilities},
   author={Kazuki Fujii and Taishi Nakamura and Mengsay Loem and Hiroki
Iida and Masanari Ohi and Kakeru Hattori and Hirai Shota and Sakae
Mizuki and Rio Yokota and Naoaki Okazaki},
   booktitle="Proceedings of the First Conference on Language Modeling",
   series={COLM},
   pages="(to appear)",
   year="2024",
   month=oct,
   address={University of Pennsylvania, USA},
}

@inproceedings{Okazaki:COLM2024,
   title={Building a Large Japanese Web Corpus for Large Language Models},
   author={Naoaki Okazaki and Kakeru Hattori and Hirai Shota and Hiroki
Iida and Masanari Ohi and Kazuki Fujii and Taishi Nakamura and Mengsay
Loem and Rio Yokota and Sakae Mizuki},
   booktitle="Proceedings of the First Conference on Language Modeling",
   series={COLM},
   pages="(to appear)",
   year="2024",
   month=oct,
   address={University of Pennsylvania, USA},
}

@misc{ma:arxiv2025,
      title={Building Instruction-Tuning Datasets from Human-Written Instructions with Open-Weight Large Language Models}, 
      author={Youmi Ma and Sakae Mizuki and Kazuki Fujii and Taishi Nakamura and Masanari Ohi and Hinari Shimada and Taihei Shiotani and Koshiro Saito and Koki Maeda and Kakeru Hattori and Takumi Okamoto and Shigeki Ishida and Rio Yokota and Hiroya Takamura and Naoaki Okazaki},
      year={2025},
      eprint={2503.23714},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2503.23714}, 
}

References

@misc{dubey2024llama3herdmodels,
      title={The Llama 3 Herd of Models}, 
      author={Abhimanyu Dubey and Abhinav Jauhri and Abhinav Pandey and Abhishek Kadian and Ahmad Al-Dahle and Aiesha Letman and Akhil Mathur and Alan Schelten and Amy Yang and Angela Fan et al.},
      year={2024},
      eprint={2407.21783},
      archivePrefix={arXiv},
      primaryClass={cs.AI},
      url={https://arxiv.org/abs/2407.21783}, 
}

⚠️ Important Note

The models released here are still in the early stages of our research and development and have not been tuned to ensure outputs align with human intent and safety considerations.

💡 Usage Tip

We thank Meta Research for releasing Llama 3.1 under a generous open license. We also received various supports including AIST, NEDO, MEXT projects and AIST program.