👑 Llama-3-Open-Ko-Linear-8B-GGUF

Quantized by llama.cpp

🏝️ Merge Details

"I thought about it yesterday—merging the solid foundation of beomi/Llama-3-Open-Ko-8B with the specialized precision of beomi/Llama-3-Open-Ko-8B-Instruct-preview, using task arithmetic, is like composing a korean song that seamlessly blends timeless rhythms with contemporary solos, creating a harmonious masterpiece tailored to today's needs."

🇰🇷 Merge Method

This model was merged using the task arithmetic merge method using beomi/Llama-3-Open-Ko-8B as a base.

🇰🇷 Models Merged

The following models were included in the merge:

• beomi/Llama-3-Open-Ko-8B-Instruct-preview

📓 Ollama

ollama create Llama-3-Open-Ko-Linear-8B -f ./Modelfile_Q5_K_M

Change it to suit your taste.

[Modelfile_Q5_K_M]

FROM llama-3-open-ko-linear-8b-Q5_K_M.gguf
TEMPLATE """
{{- if .System }}
system
<s>{{ .System }}</s>
{{- end }}
user
<s>Human:
{{ .Prompt }}</s>
assistant
<s>Assistant:
"""

SYSTEM """
친절한 챗봇으로서 상대방의 요청에 최대한 자세하고 친절하게 답하자. 모든 대답은 한국어(Korean)으로 대답해줘.
"""

PARAMETER temperature 0.7
PARAMETER num_predict 3000
PARAMETER num_ctx 4096
PARAMETER stop "<s>"
PARAMETER stop "</s>"
PARAMETER top_k 50
PARAMETER top_p 0.95

💾 Configuration

The following YAML configuration was used to produce this model:

models:
  - layer_range: [0, 31]
    model: beomi/Llama-3-Open-Ko-8B
    parameters:
      weight: 0.2
  - layer_range: [0, 31]
    model: beomi/Llama-3-Open-Ko-8B-Instruct-preview
    parameters:
      weight: 0.8
merge_method: task_arithmetic
base_model: beomi/Llama-3-Open-Ko-8B
dtype: bfloat16
random_seed: 0