This is Roformer model pretrained on texts in the Japanese language.
3.45GB wikipedia text
trained 125M step
use the BERT BPE tokenizer.
If you want to fine-tune model. Please use
from transformers import BertTokenizer, RoFormerModel
BertTokenizer.from_pretrained('Roformer-base-japanese')
RoFormerModel.from_pretrained('Roformer-base-japanese')
The accuracy in JGLUE-marc-ja-v1.0 binary sentiment classification 95.12%
Contribute by Yokohama Nationaly University Mori Lab
@article{su2021roformer, title={Roformer: Enhanced transformer with rotary position embedding}, author={Su, Jianlin and Lu, Yu and Pan, Shengfeng and Wen, Bo and Liu, Yunfeng}, journal={arXiv preprint arXiv:2104.09864}, year={2021} }