最終更新:2023-03-27 (月) 06:59:53 (389d)  

japanese-pretrained-models
Top / japanese-pretrained-models

previously: japanese-gpt2

the code for training Japanese pretrained models.

https://github.com/rinnakk/japanese-pretrained-models

モデル

language model# params# layers# emb dim# epochsdev ppltraining time*
japanese-gpt-1b?1.3B24204810+13.9n/a**
japanese-gpt2-medium?336M24102441845 days
japanese-gpt2-small?110M1276832115 days
japanese-gpt2-xsmall?37M65123284 days