README.md

---
thumbnail: https://github.com/rinnakk/japanese-pretrained-models/blob/master/rinna.png
license: gemma
datasets:
- mc4
- wikipedia
- EleutherAI/pile
- oscar-corpus/colossal-oscar-1.0
- cc100
language:
- ja
- en
tags:
- gemma2
inference: false
base_model: google/gemma-2-2b
pipeline_tag: text-generation
library_name: transformers
---

# `Gemma 2 Baku 2B (rinna/gemma-2-baku-2b)`

![rinna-icon](./rinna.png)

# Overview

We conduct continual pre-training of [google/gemma-2-2b](https://huggingface.co/google/gemma-2-2b) on **80B** tokens from a mixture of Japanese and English datasets. The continual pre-training improves the model's performance on Japanese tasks.

The name `baku` comes from the Japanese word [`獏/ばく/Baku`](https://ja.wikipedia.org/wiki/獏), which is a kind of Japanese mythical creature ([`妖怪/ようかい/Youkai`](https://ja.wikipedia.org/wiki/%E5%A6%96%E6%80%AA)).

| Size | Continual Pre-Training | Instruction-Tuning |
| :-   | :-                     | :-                 |
| 2B   | Gemma 2 Baku 2B [[HF]](https://huggingface.co/rinna/gemma-2-baku-2b) | Gemma 2 Baku 2B Instruct [[HF]](https://huggingface.co/rinna/gemma-2-baku-2b-it) |

* **Library**

    The model was trained using code based on [Lightning-AI/litgpt](https://github.com/Lightning-AI/litgpt).

* **Model architecture**

    A 26-layer, 2304-hidden-size transformer-based language model. Please refer to the [Gemma 2 Model Card](https://www.kaggle.com/models/google/gemma-2/) for detailed information on the model's architecture.

* **Training**

    The model was initialized with the [google/gemma-2-2b](https://huggingface.co/google/gemma-2-2b) model and continually trained on around **80B** tokens from a mixture of the following corpora
    - [Japanese CC-100](https://huggingface.co/datasets/cc100)
    - [Japanese C4](https://huggingface.co/datasets/mc4)
    - [Japanese OSCAR](https://huggingface.co/datasets/oscar-corpus/colossal-oscar-1.0)
    - [The Pile](https://huggingface.co/datasets/EleutherAI/pile)
    - [Wikipedia](https://dumps.wikimedia.org/other/cirrussearch)
    - rinna curated Japanese dataset
  
* **Contributors**
    - [Toshiaki Wakatsuki](https://huggingface.co/t-w)
    - [Xinqi Chen](https://huggingface.co/Keely0419)
    - [Kei Sawada](https://huggingface.co/keisawada)

---

# Benchmarking

Please refer to [rinna's LM benchmark page](https://rinnakk.github.io/research/benchmarks/lm/index.html).

---

# How to use the model

~~~python
import transformers
import torch

model_id = "rinna/gemma-2-baku-2b"
pipeline = transformers.pipeline(
    "text-generation",
    model=model_id,
    model_kwargs={"torch_dtype": torch.bfloat16, "attn_implementation": "eager"},
    device_map="auto"
)
output = pipeline(
    "西田幾多郎は、",
    max_new_tokens=256,
    do_sample=True
)
print(output[0]["generated_text"])
~~~

It is recommended to use eager attention when conducting batch inference under bfloat16 precision. 
Currently, Gemma 2 yields NaN values for input sequences with padding when the default attention mechanism (torch.scaled_dot_product_attention) is employed in conjunction with bfloat16.

---

# Tokenization
The model uses the original [google/gemma-2-2b](https://huggingface.co/google/gemma-2-2b) tokenizer.

---

# How to cite
```bibtex
@misc{rinna-gemma-2-baku-2b,
    title = {rinna/gemma-2-baku-2b},
    author = {Wakatsuki, Toshiaki and Chen, Xinqi and Sawada, Kei},
    url = {https://huggingface.co/rinna/gemma-2-baku-2b}
}

@inproceedings{sawada2024release,
    title = {Release of Pre-Trained Models for the {J}apanese Language},
    author = {Sawada, Kei and Zhao, Tianyu and Shing, Makoto and Mitsui, Kentaro and Kaga, Akio and Hono, Yukiya and Wakatsuki, Toshiaki and Mitsuda, Koh},
    booktitle = {Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024)},
    month = {5},
    year = {2024},
    pages = {13898--13905},
    url = {https://aclanthology.org/2024.lrec-main.1213},
    note = {\url{https://arxiv.org/abs/2404.01657}}
}
```
---

# References
```bibtex
@article{gemma-2-2024,
    title = {Gemma 2},
    url = {https://www.kaggle.com/models/google/gemma-2},
    publisher = {Kaggle},
    author = {Gemma Team},
    year = {2024}
}

@misc{litgpt-2023,
    author = {Lightning AI},
    title = {LitGPT},
    howpublished = {\url{https://github.com/Lightning-AI/litgpt}},
    year = {2023}
}
```
---

# License
[Gemma Terms of Use](https://ai.google.dev/gemma/terms)