README.md

metadata

license: apache-2.0

Model

• base_model : yanolja/KoSOLAR-10.7B-v0.2
• training objective: freeze, instruction Tuning

Dataset

공개 데이터 수집

• Deduplicating Training Data Makes Language Models Better 알고리즘 활용
• instruction version 1.4

Code

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

model_name = "jjingyeom/freeze_KoSoLAR-10.7B-v0.2_1.4_dedup"
model = AutoModelForCausalLM.from_pretrained(
        model_name,
)
tokenizer = AutoTokenizer.from_pretrained(model_name)