README.md

metadata

license: apache-2.0
model-index:
  - name: SOLAR_KO_1.3_deup
    results:
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: AI2 Reasoning Challenge (25-Shot)
          type: ai2_arc
          config: ARC-Challenge
          split: test
          args:
            num_few_shot: 25
        metrics:
          - type: acc_norm
            value: 55.97
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co./spaces/HuggingFaceH4/open_llm_leaderboard?query=jingyeom/SOLAR_KO_1.3_deup
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: HellaSwag (10-Shot)
          type: hellaswag
          split: validation
          args:
            num_few_shot: 10
        metrics:
          - type: acc_norm
            value: 79.97
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co./spaces/HuggingFaceH4/open_llm_leaderboard?query=jingyeom/SOLAR_KO_1.3_deup
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: MMLU (5-Shot)
          type: cais/mmlu
          config: all
          split: test
          args:
            num_few_shot: 5
        metrics:
          - type: acc
            value: 55.88
            name: accuracy
        source:
          url: >-
            https://huggingface.co./spaces/HuggingFaceH4/open_llm_leaderboard?query=jingyeom/SOLAR_KO_1.3_deup
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: TruthfulQA (0-shot)
          type: truthful_qa
          config: multiple_choice
          split: validation
          args:
            num_few_shot: 0
        metrics:
          - type: mc2
            value: 47.55
        source:
          url: >-
            https://huggingface.co./spaces/HuggingFaceH4/open_llm_leaderboard?query=jingyeom/SOLAR_KO_1.3_deup
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: Winogrande (5-shot)
          type: winogrande
          config: winogrande_xl
          split: validation
          args:
            num_few_shot: 5
        metrics:
          - type: acc
            value: 76.87
            name: accuracy
        source:
          url: >-
            https://huggingface.co./spaces/HuggingFaceH4/open_llm_leaderboard?query=jingyeom/SOLAR_KO_1.3_deup
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: GSM8k (5-shot)
          type: gsm8k
          config: main
          split: test
          args:
            num_few_shot: 5
        metrics:
          - type: acc
            value: 22.59
            name: accuracy
        source:
          url: >-
            https://huggingface.co./spaces/HuggingFaceH4/open_llm_leaderboard?query=jingyeom/SOLAR_KO_1.3_deup
          name: Open LLM Leaderboard

Model

base_model : beomi/OPEN-SOLAR-KO-10.7B

Dataset

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

Code

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

model_name = "jingyeom/SOLAR_KO_1.3_deup"
model = AutoModelForCausalLM.from_pretrained(
        model_name,
)
tokenizer = AutoTokenizer.from_pretrained(model_name)

Benchmark

Ko-LLM-Leaderboard (24.01.29 기준 리더보드 11등)

Average	Ko-ARC	Ko-HellaSwag	Ko-MMLU	Ko-TruthfulQA	Ko-CommonGen V2
53.63	52.65	60.92	50.9	45.14	58.56

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	56.47
AI2 Reasoning Challenge (25-Shot)	55.97
HellaSwag (10-Shot)	79.97
MMLU (5-Shot)	55.88
TruthfulQA (0-shot)	47.55
Winogrande (5-shot)	76.87
GSM8k (5-shot)	22.59