POLAR is a Korean LLM developed by Plateer's AI-lab. It was inspired by Upstage's SOLAR. We will continue to evolve this model and hope to contribute to the Korean LLM ecosystem.
Developed by: AI-Lab of Plateer(Woomun Jung, Eunsoo Ha, MinYoung Joo, Seongjun Son)
Model type: Language model
Language(s) (NLP): ko
License: apache-2.0
Parent Model: x2bee/POLAR-14B-v0.2
wiki Data 를 활용하여 DPO 데이터 1.4 만건 생성하여 학습했습니다.
Direct Use
from transformers import AutoTokenizer, AutoModelForCausalLM
tokenizer = AutoTokenizer.from_pretrained("x2bee/POLAR-14B-DPO-v1.4")
model = AutoModelForCausalLM.from_pretrained("x2bee/POLAR-14B-DPO-v1.4")
Downstream Use [Optional]
Out-of-Scope Use
Bias, Risks, and Limitations
Significant research has explored bias and fairness issues with language models (see, e.g., Sheng et al. (2021) and Bender et al. (2021)). Predictions generated by the model may include disturbing and harmful stereotypes across protected classes; identity characteristics; and sensitive, social, and occupational groups.
