README.md

metadata

license: apache-2.0
model-index:
  - name: freeze_KoSoLAR-10.7B-v0.2_1.4_dedup
    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: 58.45
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=jingyeom/freeze_KoSoLAR-10.7B-v0.2_1.4_dedup
          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: 81.26
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=jingyeom/freeze_KoSoLAR-10.7B-v0.2_1.4_dedup
          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: 64.83
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=jingyeom/freeze_KoSoLAR-10.7B-v0.2_1.4_dedup
          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: 44.5
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=jingyeom/freeze_KoSoLAR-10.7B-v0.2_1.4_dedup
          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: 79.08
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=jingyeom/freeze_KoSoLAR-10.7B-v0.2_1.4_dedup
          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: 32.22
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=jingyeom/freeze_KoSoLAR-10.7B-v0.2_1.4_dedup
          name: Open LLM Leaderboard

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)

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	60.06
AI2 Reasoning Challenge (25-Shot)	58.45
HellaSwag (10-Shot)	81.26
MMLU (5-Shot)	64.83
TruthfulQA (0-shot)	44.50
Winogrande (5-shot)	79.08
GSM8k (5-shot)	32.22