---
license: apache-2.0
language:
  - en
  - ja
programming_language:
  - C
  - C++
  - C#
  - Go
  - Java
  - JavaScript
  - Lua
  - PHP
  - Python
  - Ruby
  - Rust
  - Scala
  - TypeScript
pipeline_tag: text-generation
inference: false
---

# llm-jp-13b-v1.0-mdsfmt

This repository provides large language models (Megatron-DeepSpeed format) developed by [LLM-jp](https://llm-jp.nii.ac.jp/), a collaborative project launched in Japan. **Hugging Face Transformers format models are available [here](https://huggingface.co/llm-jp).**

| Model Variant |
| :--- |
|**Pre-trained models** <span style="color:red">(Megatron-DeepSpeed format)</span>|
| [llm-jp-13b-v1.0-mdsfmt](https://huggingface.co/llm-jp/llm-jp-13b-v1.0-mdsfmt) |
| [llm-jp-13b-v1.0-mdsfmt-itr87870](https://huggingface.co/llm-jp/llm-jp-13b-v1.0-mdsfmt-itr87870) |
| [llm-jp-1.3b-v1.0-mdsfmt](https://huggingface.co/llm-jp/llm-jp-1.3b-v1.0-mdsfmt) |
| [llm-jp-1.3b-v1.0-mdsfmt-itr87430](https://huggingface.co/llm-jp/llm-jp-1.3b-v1.0-mdsfmt-itr87430) | 


`llm-jp-13b-v1.0-mdsfmt-itr87870` 
and `llm-jp-1.3b-v1.0-mdsfmt-itr87430` 
were originally trained with approximately 270B+ tokens. 
`llm-jp-13b-v1.0-mdsfmt` 
and `llm-jp-1.3b-v1.0-mdsfmt` 
are models further trained by additional (potentially) high-quality 27B tokens data from `llm-jp-13b-v1.0-mdsfmt-itr87870` and `llm-jp-1.3b-v1.0-mdsfmt-itr87430`, respectively for finalizing the pre-training.


## Model Details

- **Model type:** Transformer-based Language Model
- **Total seen tokens:** 300B

|Model|Params|Layers|Hidden size|Heads|Context length|
|:---:|:---:|:---:|:---:|:---:|:---:|
|13b model|13b|40|5120|40|2048|
|1.3b model|1.3b|24|2048|16|2048|


## Training

- **Pre-training:**
  - **Hardware:** 96 A100 40GB GPUs ([mdx cluster](https://mdx.jp/en/))
  - **Software:** Megatron-DeepSpeed


## Tokenizer
The tokenizer of this model is based on [huggingface/tokenizers](https://github.com/huggingface/tokenizers) Unigram byte-fallback model.
The vocabulary entries were converted from [`llm-jp-tokenizer v2.1 (50k)`](https://github.com/llm-jp/llm-jp-tokenizer/releases/tag/v2.1).
Please refer to [README.md](https://github.com/llm-jp/llm-jp-tokenizer) of `llm-ja-tokenizer` for details on the vocabulary construction procedure.
- **Model:** Hugging Face Fast Tokenizer using Unigram byte-fallback model which requires `tokenizers>=0.14.0`
- **Training algorithm:** SentencePiece Unigram byte-fallback
- **Training data:** A subset of the datasets for model pre-training
- **Vocabulary size:** 50,570 (mixed vocabulary of Japanese, English, and source code)


## Datasets

### Pre-training

The models have been pre-trained using a blend of the following datasets.

| Language | Dataset | Tokens |
|:---:|:---:|:---:|
|Japanese|[Wikipedia](https://huggingface.co/datasets/wikipedia)|1.5B
||[mC4](https://huggingface.co/datasets/mc4)|136B
|English|[Wikipedia](https://huggingface.co/datasets/wikipedia)|5B
||[The Pile](https://huggingface.co/datasets/EleutherAI/pile)|135B
|Codes|[The Stack](https://huggingface.co/datasets/bigcode/the-stack)|10B

The pre-training was continuously conducted using a total of 10 folds of non-overlapping data, each consisting of approximately 27-28B tokens.
We finalized the pre-training with additional (potentially) high-quality 27B tokens data obtained from the identical source datasets listed above used for the 10-fold data.


## Evaluation

You can view the evaluation results of several LLMs on this [leaderboard](http://wandb.me/llm-jp-leaderboard). We used [llm-jp-eval](https://github.com/llm-jp/llm-jp-eval) for the evaluation.


## Risks and Limitations

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.


## Send Questions to

llm-jp(at)nii.ac.jp


## License

[Apache License, Version 2.0](https://www.apache.org/licenses/LICENSE-2.0)


## Model Card Authors
*The names are listed in alphabetical order.*

Hirokazu Kiyomaru, Hiroshi Matsuda, Jun Suzuki, Namgi Han, Saku Sugawara, Shota Sasaki, Shuhei Kurita, Taishi Nakamura, Takumi Okamoto.