metadata

base_model:
  - davidkim205/nox-solar-10.7b-v2
  - chihoonlee10/T3Q-ko-solar-dpo-v6.0
library_name: transformers
tags:
  - mergekit
  - merge

model_storage

This is a merge of pre-trained language models created using mergekit.

Merge Details

Merge Method

This model was merged using the SLERP merge method.

Models Merged

The following models were included in the merge:

Configuration

The following YAML configuration was used to produce this model:

base_model:
  model:
    path: chihoonlee10/T3Q-ko-solar-dpo-v6.0
dtype: float16
merge_method: slerp
parameters:
  t:
  - filter: self_attn
    value: [0.0, 0.5, 0.3, 0.7, 1.0]
  - filter: mlp
    value: [1.0, 0.5, 0.7, 0.3, 0.0]
  - value: 0.5
slices:
- sources:
  - layer_range: [0, 47]
    model:
      model:
        path: chihoonlee10/T3Q-ko-solar-dpo-v6.0
  - layer_range: [0, 47]
    model:
      model:
        path: davidkim205/nox-solar-10.7b-v2

Evaluation Results

모델 명칭	Average	HellaSwag	COPA	BooIQ
KoGPT	58.2	55.9	73.5	45.1
Polyglot-ko-13B	62.4	59.5	79.4	48.2
LLaMA 2-13B	45.2	41.3	59.3	34.9
Baichuan 2-13B	52.7	39.2	60.6	58.4
QWEN-14B	47.8	45.3	64.9	33.4
Orion-14B-Chat	68.8	47.0	77.7	81.6
Ocelot-ko-10.8B	72.5	50.0	75.8	91.7

Evaluation Results

Model	글쓰기	이해	문법
gpt-3.5-turbo-0125	8.78	9.57	6.50
HyperClovaX	8.50	9.50	8.50
allganize/Llama-3-Alpha-Ko-8B-Instruct	8.50	8.35	4.92
Synatra-kiqu-7B	4.42	5.71	4.50
Ocelot-ko-10.8B	8.57	7.00	6.57