---
license: llama3
language:
- en
- ja
metrics:
- comet
pipeline_tag: translation
tags:
- machine translation
- MT
- llama-3
---

# Overview
The model is based on rinna's [rinna/llama-3-youko-8b], fine-tuned using LoRA on a small number of parallel sentences from English to Japanese. The model has a COMET (Unbabel/wmt22-comet-da) of 0.9011 on flores200 devtest.

* **Model architecture**

    A 32-layer, 4096-hidden-size transformer-based language model. Refer to the [Llama 3 Model Card](https://github.com/meta-llama/llama3/blob/main/MODEL_CARD.md) for architecture details.
* **Training: Built with Meta Llama 3**

    The model was initialized with the [meta-llama/Meta-Llama-3-8B](https://huggingface.co/meta-llama/Meta-Llama-3-8B) model and continually trained on around **22B** 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**
    - [Koh Mitsuda](https://huggingface.co/mitsu-koh)
    - [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 torch
from transformers import AutoModelForCausalLM, AutoTokenizer

response_template = "\n###  日本語：\n"
prefix = "###  次の英語の文書を日本語に翻訳してください：\n"


def create_input(text, tokenizer):
    text = f"{prefix}{text}{response_template}"
    input_ids = tokenizer.encode(text, return_tensors="pt")
    return input_ids


model_id = "lyu/MT/output/llama3-sft-lora-16-NLLB-100k-run2/merge"
model = AutoModelForCausalLM.from_pretrained(
    model_id, torch_dtype=torch.bfloat16, attn_implementation="flash_attention_2"
).cuda()
tokenizer = AutoTokenizer.from_pretrained(model_id, use_fast=True)

en = "LLMs Are Here but Not Quite There Yet"
input_ids = create_input(en, tokenizer).to(model.device)
outputs = model.generate(
    input_ids,
    max_new_tokens=256,
    num_beams=5,
    do_sample=False,
    early_stopping=True,
)
response = outputs[0][input_ids.shape[-1] :]
print(tokenizer.decode(response, skip_special_tokens=True))
~~~~

---

# Tokenization
The model uses the original meta-llama/Meta-Llama-3-8B tokenizer.


# References
```bibtex
@article{llama3modelcard,
    title={Llama 3 Model Card},
    author={AI@Meta},
    year={2024},
    url = {https://github.com/meta-llama/llama3/blob/main/MODEL_CARD.md}
}
@software{gpt-neox-library,
    title = {{GPT-NeoX: Large Scale Autoregressive Language Modeling in PyTorch}},
    author = {Andonian, Alex and Anthony, Quentin and Biderman, Stella and Black, Sid and Gali, Preetham and Gao, Leo and Hallahan, Eric and Levy-Kramer, Josh and Leahy, Connor and Nestler, Lucas and Parker, Kip and Pieler, Michael and Purohit, Shivanshu and Songz, Tri and Phil, Wang and Weinbach, Samuel},
    doi = {10.5281/zenodo.5879544},
    month = {8},
    year = {2021},
    version = {0.0.1},
    url = {https://www.github.com/eleutherai/gpt-neox},
}
@misc{rinna-llama-3-youko-8b, 
    title = {rinna/llama-3-youko-8b}, 
    author = {Mitsuda, Koh and Sawada, Kei},
    url = {https://huggingface.co/rinna/llama-3-youko-8b}, 
}
@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},
    url = {https://arxiv.org/abs/2404.01657},
}
```
---

# License