---
license: apache-2.0
datasets:
- SkelterLabsInc/JaQuAD
language:
- ja
---

MambaSan-370m 🐍

MambaSan-370m is the first chat Japanese language model based on a state-space model architecture (Mamba).

The model is based on Albert Gu's and Tri Dao's work Mamba: Linear-Time Sequence Modeling with Selective State Spaces (paper) as well as their model implementation. .

The Code used for pretraining will soon be published on my github: https://github.com/lcabannes
Citation

bibtex
@misc{lcabannes2024MambaSan-370m,
  title        = {MambaSan-370m},
  author       = {Loïc Cabannes},
  year         = {2024},
  howpublished = {HuggingFace},
  url          = {https://huggingface.co/loiccabannes/MambaSan-370m/}
}