Korean-PLM (Korean Pre-trained Language Models)
공개된 한국어 사전학습 모델을 기록합니다. 크게 아래 3개의 모델 계열로 구분했으며, 모델 사이즈는 정확하지 않을 수 있습니다.
- Encoder Model (BERT 계열)
- Decoder Model (GPT 계열)
- Encoder-Decoder Model (Seq2seq 계열)
Model Name |
Size |
Link |
LASSL BERT |
Small (10M) |
link |
LASSL RoBERTa |
Small (10M) |
link |
TUNiB ELECTRA (Ko) |
Small (10M) |
link |
TUNiB ELECTRA (Ko-En) |
Small (10M) |
link |
KoELECTRA |
Small (10M) |
link |
KcELECTRA |
Small (10M) |
link |
Soongsil-BERT |
Small (10M) |
link |
Ko-CHAR-ELECTRA |
Small (10M) |
link |
Distill-KoBERT |
30M |
link |
KoBERT |
Base (110M) |
link |
KorBERT |
Base (110M) |
link |
HanBERT |
Base (110M) |
link |
KcBERT |
Base (110M) |
link |
Soongsil-BERT |
Base (110M) |
link |
KLUE-RoBERTa |
Base (110M) |
link |
KoELECTRA |
Base (110M) |
link |
KcELECTRA |
Base (110M) |
link |
TUNiB-ELECTRA (Ko) |
Base (110M) |
link |
TUNiB-ELECTRA (Ko-En) |
Base (110M) |
link |
LMKor Albert |
Base (110M) |
link |
LMKor Bert |
Base (110M) |
link |
LMKor Funnel |
Base (110M) |
link |
LMKor ELECTRA |
Base (110M) |
link |
KalBERT |
Base (110M) |
link |
LASSL BERT |
Base (110M) |
link |
KoBigBird |
Base (110M) |
link |
Ko-CHAR-BERT |
Base (110M) |
link |
Ko-CHAR-ELECTRA |
Base (110M) |
link |
BrainSBERT |
Base (110M) |
link |
KPFBERT |
Base (110M) |
link |
KcBERT |
Large (340M) |
link |
KLUE-RoBERTa |
Large (340M) |
link |
Model Name |
Size |
Link |
LASSL GPT2 |
Small (20M) |
link |
LMKor KoGPT2 |
Base (110M) |
link |
SKT KoGPT2 |
Base (110M) |
link |
SKT KoGPT Trinity |
1.2B |
link |
KakaoBrain KoGPT |
6B |
link |
EleutherAI Polyglot-Ko |
1.3B |
link |
EleutherAI Polyglot-Ko |
3.8B |
link |
EleutherAI Polyglot-Ko |
5.8B |
link |
EleutherAI Polyglot-Ko |
12.8B |
link |
KoAlpaca (Polyglot-Ko) |
5.8B |
link |
KoAlpaca (Polyglot-Ko) |
12.8B |
link |
Encoder-Decoder Model (Seq2seq 계열)
Model Name |
Size |
Link |
Cosmoquester BART |
Mini (13M) |
link |
KE-T5 (Ko-En) |
Small (60M) |
link |
KE-T5 (Ko) |
Small (60M) |
link |
LMKor T5 |
Small (60M) |
link |
Cosmoquester BART |
Small (40M) |
link |
LMKor BERT Shared |
Base (130M) |
link |
SKT KoBART |
Base (130M) |
link |
Cosmoquester BART |
Base (130M) |
link |
Kolang-T5 |
Base (220M) |
link |
KcT5 |
Base (220M) |
link |
KE-T5 (Ko-En) |
Base (220M) |
link |
KE-T5 (Ko) |
Base (220M) |
link |
Asian BART (Ko) |
Base (360M) |
link |
pko-t5 (Ko) |
Base (250M) |
link |
Wisenut KoT5 (Ko) |
Base (220M) |
link |
mBART |
Large (610M) |
link |
KE-T5 (Ko-En) |
Large (770M) |
link |
KE-T5 (Ko) |
Large (770M) |
link |
pko-t5 (Ko) |
Large (800M) |
link |