README.md

---
license: apache-2.0
---
# RakutenAI-7B
## Model Description
RakutenAI-7B is a systematic initiative that brings the latest technologies to the world of Japanese LLMs. RakutenAI-7B achieves the best scores on the Japanese language understanding benchmarks while maintaining a competitive performance on the English test sets among similar models such as OpenCalm, Elyza, Youri, Nekomata and Swallow. RakutenAI-7B leverages the Mistral model architecture and is based on [Mistral-7B-v0.1](https://huggingface.co/mistralai/Mistral-7B-v0.1) pre-trained checkpoint, exemplifying a successful retrofitting of the pre-trained model weights. Moreover, we extend Mistral's vocabulary from 32k to 48k to offer a better character-per-token rate for Japanese.

*If you are looking for an instruction-tuned model, check [RakutenAI-7B-instruct](https://huggingface.co/Rakuten/RakutenAI-7B-instruct)*.

*If you are looking for a chat-tuned model, check [RakutenAI-7B-chat](https://huggingface.co/Rakuten/RakutenAI-7B-chat)*.

## Usage

```python
from transformers import AutoModelForCausalLM, AutoTokenizer

model_path = "Rakuten/RakutenAI-7B"
tokenizer = AutoTokenizer.from_pretrained(model_path)
model = AutoModelForCausalLM.from_pretrained(model_path, torch_dtype="auto", device_map="auto")
model.eval()

requests = [
    "南硫黄島原生自然環境保全地域は、自然",
    "The capybara is a giant cavy rodent",
]

for req in requests:
    input_ids = tokenizer.encode(req, return_tensors="pt").to(device=model.device)
    tokens = model.generate(
        input_ids,
        max_new_tokens=256,
        do_sample=True,
        repetition_penalty=1.1,
        pad_token_id=tokenizer.eos_token_id,
    )
    out = tokenizer.decode(tokens[0], skip_special_tokens=True)
    print("INPUT:\n" + req)
    print("OUTPUT:\n" + out)
    print()
    print()
```

## Model Details

* **Developed by**: [Rakuten Group, Inc.](https://ai.rakuten.com/)
* **Language(s)**: Japanese, English
* **License**: This model is licensed under [Apache License, Version 2.0](https://www.apache.org/licenses/LICENSE-2.0).

### Limitations and Bias

The suite of RakutenAI-7B models is capable of generating human-like text on a wide range of topics. However, like all LLMs, they have limitations and can produce biased, inaccurate, or unsafe outputs. Please exercise caution and judgement while interacting with them.

## Citation
For citing our work on the suite of RakutenAI-7B models, please use: 

```
@misc{2024RakutenAI-7B,
      title={RakutenAI-7B: Extending Large Language Models for Japanese}, 
      author={Rakuten Group, Inc.},
      year={2024},
      eprint={},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}
```