jingyeom/KoSoLAR-10.7B-v0.2_1.4_dedup

Model

base_model : yanolja/KoSOLAR-10.7B-v0.2

Dataset

• 공개 데이터 수집
• Deduplicating Training Data Makes Language Models Better 알고리즘 활용

Code

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

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

Benchmark

Ko-LLM-Leaderboard