zomd_-_AISquare-Instruct-yi-ko-6b-v0.9.30-4bits

Quantization made by Richard Erkhov.

Github

Discord

Request more models

AISquare-Instruct-yi-ko-6b-v0.9.30 - bnb 4bits

• Model creator: https://huggingface.co/zomd/
• Original model: https://huggingface.co/zomd/AISquare-Instruct-yi-ko-6b-v0.9.30/

Original model description:

language:

• en pipeline_tag: text-generation license: cc-by-nc-4.0

---

AISquare-Instruct-yi-ko-6b-v0.9.30

Model Details

Developed by Inswave Systems UI Platform Team

Method
Using DPO method and SFT method

Hardware
We utilized an A100x4 * 1 for training our model

Base Model beomi/Yi-Ko-6B

Open ko-leaderboard Rank

<img src='./ko-leaderboard.png' width=512>

Implementation Code

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

repo = "zomd/AISquare-Instruct-yi-ko-6b-v0.9.30"
model = AutoModelForCausalLM.from_pretrained(
        repo,
        return_dict=True,
        torch_dtype=torch.float16,
        device_map='auto'
)
tokenizer = AutoTokenizer.from_pretrained(repo)

이 연구는 인공지능산업융합사업단(AICA)에서 추진한 『인공지능 중심 산업융합 집적단지 조성사업』 의 지원을 받아 진행된 결과입니다.