|
--- |
|
language: |
|
- ja |
|
- en |
|
license: other |
|
license_link: LICENSE |
|
--- |
|
|
|
# Sarashina2-8x70B |
|
|
|
This repository provides large language models trained by [SB Intuitions](https://www.sbintuitions.co.jp/). |
|
|
|
|
|
## Required Hardware |
|
BF16 Inference: |
|
- 16x H100 |
|
- 16x A100 80GB |
|
|
|
|
|
## Model Description |
|
|
|
We constructed this Sarashina2-8x70B model, which consists of over 450 billion parameters, by applying the [sparse upcycling technique](https://arxiv.org/abs/2212.05055) to our [Sarashina2-70B](https://huggingface.co/sbintuitions/sarashina2-70b) model to efficiently build the Mixture-of-Experts model. |
|
We trained the Sarashina2-8x70B model using a mix of Japanese and English corpora from web data. |
|
|
|
|
|
## Tokenization |
|
|
|
We use a [sentencepiece](https://github.com/google/sentencepiece) tokenizer with a unigram language model and byte-fallback. |
|
We do not apply pre-tokenization with Japanese tokenizer. |
|
Thus, a user may directly feed raw sentences into the tokenizer. |
|
|
|
|
|
## Ethical Considerations and Limitations |
|
Sarashina2 has not been tuned to follow an instruction yet. |
|
Therefore, sarashina2 might generate some meaningless sequences, some inaccurate instances or biased/objectionable outputs. |
|
Before using sarashina2, we would like developers to tune models based on human preferences and safety considerations. |
|
|
|
## License |
|
|
|
[Sarashina Model NonCommercial License Agreement](https://huggingface.co/sbintuitions/sarashina2-8x70B/blob/main/Sarashina%20Model%20NonCommercial%20License%20Agreement) |
