README.md

metadata

language:
  - ko
license: cc-by-nc-4.0
tags:
  - merge
  - lazymergekit
  - LDCC/LDCC-SOLAR-10.7B
  - upstage/SOLAR-10.7B-Instruct-v1.0
base_model:
  - LDCC/LDCC-SOLAR-10.7B
  - upstage/SOLAR-10.7B-Instruct-v1.0
model-index:
  - name: SOLAR-10.7B-slerp
    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: 68.17
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=SJ-Donald/SOLAR-10.7B-slerp
          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: 86.91
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=SJ-Donald/SOLAR-10.7B-slerp
          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: 66.73
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=SJ-Donald/SOLAR-10.7B-slerp
          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: 67.42
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=SJ-Donald/SOLAR-10.7B-slerp
          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: 84.06
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=SJ-Donald/SOLAR-10.7B-slerp
          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: 62.17
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=SJ-Donald/SOLAR-10.7B-slerp
          name: Open LLM Leaderboard

SOLAR-10.7B-slerp

SOLAR-10.7B-slerp is a merge of the following models using mergekit:

• LDCC/LDCC-SOLAR-10.7B
• upstage/SOLAR-10.7B-Instruct-v1.0

Github

https://github.com/sunjin7725/SOLAR-10.7b-slerp

Benchmark

Open-Ko-LLM-Leaderboard

Average	Ko-ARC	Ko-HellaSwag	Ko-MMLU	Ko-TruthfulQA	Ko-CommonGen V2
56.93	53.58	62.03	53.31	57.16	58.56

How to use

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer

repo = 'SJ-Donald/SOLAR-10.7B-slerp'

tokenizer = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(
    repo,
    return_dict=True,
    torch_dtype=torch.float16,
    device_map='auto'
)

🧩 Configuration

slices:
  - sources:
      - model: LDCC/LDCC-SOLAR-10.7B
        layer_range: [0, 48]
      - model: upstage/SOLAR-10.7B-Instruct-v1.0
        layer_range: [0, 48]
merge_method: slerp
base_model: upstage/SOLAR-10.7B-Instruct-v1.0
parameters:
  t:
    - filter: self_attn
      value: [0, 0.5, 0.3, 0.7, 1]
    - filter: mlp
      value: [1, 0.5, 0.7, 0.3, 0]
    - value: 0.5
tokenizer_source: union
dtype: float16

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	72.58
AI2 Reasoning Challenge (25-Shot)	68.17
HellaSwag (10-Shot)	86.91
MMLU (5-Shot)	66.73
TruthfulQA (0-shot)	67.42
Winogrande (5-shot)	84.06
GSM8k (5-shot)	62.17