Skip to content

MrBananaHuman/KoreanCharacterBert

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

19 Commits
 
 
 
 
 
 

Repository files navigation

KoreanCharacterBert

Korean BERT model using character tokenizer

Small model (3 layers) and base model (12 layers) download link

KorQuAD v1.0 results

  • Small model (450,000 steps, 512 batch size): f1=86.75, em=77.14
  • Base model (450,000 steps, 128 batch size): f1=90.59, em=82.47

Inference time (RTX 2070 Super, 1 batch size)

  • Small model: 10 ms
  • Base model: 30 ms

About

Korean BERT model using character tokenizer

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages