Dear_My_best_Friends-v2-13B

The license is cc-by-nc-sa-4.0.

(주)미디어그룹사람과숲과 (주)마커의 LLM 연구 컨소시엄으로 개발된 모델입니다

🌙Dear_My_best_Friends-v2-13B🌙

The main image is generated image using playground AI.

Model Details

Model Developers Seungyoo Lee (DopeorNope)

Input Models input text only.

Output Models generate text only.

Model Architecture
Dear_My_best_Friends-13B is an auto-regressive 13B language model based on the LLaMA2 transformer architecture.

Base Model DopeorNope/Dear_My_best_Friend-SFT-v2-13B- not uploaded yet

COKAL_pre_DPO_Test_v3-13b is the SFT model to train the DPO method.

Training Dataset

• DPO training dataset: DopeorNope/DPO-Ko-Dataset - private

This dataset was constructed by directly collecting and reorganizing data by DopeorNope, obtaining insights from "lvwerra/stack-exchange-paired" to create a paired dataset. (It means I do not use stack-exchange-paired; I just got an insight from it.)

• SFT training dataset: DopeorNope/New_Data_Technology - private

This dataset is based on "HumanF-MarkrAI's private data" and has been processed using the Near Dedup algorithm to remove items with a Jaccard Similarity threshold of 0.8 or higher. In addition, inconsistent inputs have been cleaned and modified. Moreover, I implemented a new method(It is a test version, and I will share it soon).

Training
I developed the model in an environment with four RTX 3090 GPUs running Ubuntu 18.04. It seems that when uploading the model directly to a repository from a Linux server, there may be an issue causing the model to appear to have more parameters. However, this model is based on a 13B architecture.

Implementation Code

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

repo = "Dear_My_best_Friends-v2-13B"
model = AutoModelForCausalLM.from_pretrained(
        repo,
        return_dict=True,
        torch_dtype=torch.float16,
        device_map='auto'
)
model_tokenizer = AutoTokenizer.from_pretrained(repo)

Acknowledgement

이 모델은 과학기술정보통신부·광주광역시가 공동 지원한 '인공지능 중심 산업융합 집적단지 조성사업'으로 지원을 받아 수행된 연구 결과입니다.

This model was supported by Artificial intelligence industrial convergence cluster development project funded by the Ministry of Science and ICT(MSIT, Korea)&Gwangju Metropolitan City.

---