README.md

metadata

language: ko

Albert base model for Korean

• 70GB Korean text dataset and 42000 lower-cased subwords are used
• Check the model performance and other language models for Korean in github

from transformers import BertTokenizerFast, AlbertModel

tokenizer_albert = BertTokenizerFast.from_pretrained("kykim/albert-kor-base")
model_albert = AlbertModel.from_pretrained("kykim/albert-kor-base")