aeolian83's picture
Update README.md
637265b
|
raw
history blame
474 Bytes
---
license: apache-2.0
language:
- ko
datasets:
- squarelike/sharegpt_deepl_ko_translation
tags:
- pytorch
- causal-lm
---
# poly-ko-1.3b-translate
- EleutherAI/polyglot-ko-1.3b을 squarelike/sharegpt_deepl_ko_translation으로 영한 번역만 가능하도록 fine-tuning한 모델
- QRoLA기법으로 fine-tunnig
### 훈련 정보
- Epoch: 3
- learning-rate: 3e-4
- batch_size: 1
- Lora r: 8
- Lora target modules: query_key_value
3090GPU 1대로 훈련했습니다.