Instructions to use muzerai/Llama-3.1-KoEn-8B-magic8-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use muzerai/Llama-3.1-KoEn-8B-magic8-GGUF with Transformers:
# Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("muzerai/Llama-3.1-KoEn-8B-magic8-GGUF", dtype="auto") - llama-cpp-python
How to use muzerai/Llama-3.1-KoEn-8B-magic8-GGUF with llama-cpp-python:
# !pip install llama-cpp-python from llama_cpp import Llama llm = Llama.from_pretrained( repo_id="muzerai/Llama-3.1-KoEn-8B-magic8-GGUF", filename="llama-3.1-koen-8b-magic8-Q4_K_M.gguf", )
output = llm( "Once upon a time,", max_tokens=512, echo=True ) print(output)
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- llama.cpp
How to use muzerai/Llama-3.1-KoEn-8B-magic8-GGUF with llama.cpp:
Install from brew
brew install llama.cpp # Start a local OpenAI-compatible server with a web UI: llama-server -hf muzerai/Llama-3.1-KoEn-8B-magic8-GGUF:Q4_K_M # Run inference directly in the terminal: llama-cli -hf muzerai/Llama-3.1-KoEn-8B-magic8-GGUF:Q4_K_M
Install from WinGet (Windows)
winget install llama.cpp # Start a local OpenAI-compatible server with a web UI: llama-server -hf muzerai/Llama-3.1-KoEn-8B-magic8-GGUF:Q4_K_M # Run inference directly in the terminal: llama-cli -hf muzerai/Llama-3.1-KoEn-8B-magic8-GGUF:Q4_K_M
Use pre-built binary
# Download pre-built binary from: # https://github.com/ggerganov/llama.cpp/releases # Start a local OpenAI-compatible server with a web UI: ./llama-server -hf muzerai/Llama-3.1-KoEn-8B-magic8-GGUF:Q4_K_M # Run inference directly in the terminal: ./llama-cli -hf muzerai/Llama-3.1-KoEn-8B-magic8-GGUF:Q4_K_M
Build from source code
git clone https://github.com/ggerganov/llama.cpp.git cd llama.cpp cmake -B build cmake --build build -j --target llama-server llama-cli # Start a local OpenAI-compatible server with a web UI: ./build/bin/llama-server -hf muzerai/Llama-3.1-KoEn-8B-magic8-GGUF:Q4_K_M # Run inference directly in the terminal: ./build/bin/llama-cli -hf muzerai/Llama-3.1-KoEn-8B-magic8-GGUF:Q4_K_M
Use Docker
docker model run hf.co/muzerai/Llama-3.1-KoEn-8B-magic8-GGUF:Q4_K_M
- LM Studio
- Jan
- Ollama
How to use muzerai/Llama-3.1-KoEn-8B-magic8-GGUF with Ollama:
ollama run hf.co/muzerai/Llama-3.1-KoEn-8B-magic8-GGUF:Q4_K_M
- Unsloth Studio
How to use muzerai/Llama-3.1-KoEn-8B-magic8-GGUF with Unsloth Studio:
Install Unsloth Studio (macOS, Linux, WSL)
curl -fsSL https://unsloth.ai/install.sh | sh # Run unsloth studio unsloth studio -H 0.0.0.0 -p 8888 # Then open http://localhost:8888 in your browser # Search for muzerai/Llama-3.1-KoEn-8B-magic8-GGUF to start chatting
Install Unsloth Studio (Windows)
irm https://unsloth.ai/install.ps1 | iex # Run unsloth studio unsloth studio -H 0.0.0.0 -p 8888 # Then open http://localhost:8888 in your browser # Search for muzerai/Llama-3.1-KoEn-8B-magic8-GGUF to start chatting
Using HuggingFace Spaces for Unsloth
# No setup required # Open https://huggingface.co/spaces/unsloth/studio in your browser # Search for muzerai/Llama-3.1-KoEn-8B-magic8-GGUF to start chatting
- Atomic Chat new
- Docker Model Runner
How to use muzerai/Llama-3.1-KoEn-8B-magic8-GGUF with Docker Model Runner:
docker model run hf.co/muzerai/Llama-3.1-KoEn-8B-magic8-GGUF:Q4_K_M
- Lemonade
How to use muzerai/Llama-3.1-KoEn-8B-magic8-GGUF with Lemonade:
Pull the model
# Download Lemonade from https://lemonade-server.ai/ lemonade pull muzerai/Llama-3.1-KoEn-8B-magic8-GGUF:Q4_K_M
Run and chat with the model
lemonade run user.Llama-3.1-KoEn-8B-magic8-GGUF-Q4_K_M
List all available models
lemonade list
Llama-3.1-Ko-8B-magic8 made by "AIJOAH"
The merged model combining Llama-3.1-Korean-8B-Instruct and Llama-3.1-Storm-8B improves performance like โ including Korean-language instruction following, multilingual knowledge-based QA, reasoning, reduced hallucinations, and structured output generation (e.g., JSON, Markdown). This merge is particularly beneficial for developers seeking a strong Korean-capable model that also excels in logic, accuracy, and function-calling, while remaining lightweight enough for local inference environments such as Ollama or vLLM.
Merge Method
This model was merged using the DELLA
Models Merged
The following models were included in the merge:
Test Q5
ollama create modelname -f Modelfile
>>> ์ด๋ ฅ์ ์ฐ๋ ๋ฐฉ๋ฒ์ ์ค๋ช
ํด์ค
์ด๋ ฅ์ ๋๋ ๋ฆฌ์กธ๋ฃจ์
(Resume)๋, ์์ ์ ๊ฒฝ๋ ฅ์ ์์ฝํ์ฌ ์ ๋ฆฌํ ๋ฌธ์๋ก, ์ ์ง๋ฌด๋ฅผ ํฌํจํ ๊ฒฝํ๊ณผ ์๊ฒฉ์ ๋ช
ํํ๊ฒ ํํํ๋ ๋ฐ ์ฌ์ฉ๋ฉ๋๋ค. ์ด๋ ฅ์๋ ์ฃผ์ ๊ธฐ์
์ด๋ ์กฐ์ง์ด ์ง์
์๋ฅผ ํ๊ฐํ๊ณ ์ต์ข
์ ์ผ๋ก ์ต์ ์ ์ธ์ฌ๋ฅผ ์ ๋ฐํ๊ธฐ ์ํด ํ์์ ์ธ ๋๊ตฌ์
๋๋ค.
1. ์ด๋ ฅ์ ์์ฑ์ ํ์ํ ์ ๋ณด:
- ์ด๋ฆ ๋ฐ ์ฐ๋ฝ์ฒ
- ์ง์
์ ๋ชฉ ๋ฐ ์ง์
- ํ์ฌ/์ ์ง๋ฌด ๋ฐ ํ์ฌ๋ช
(์ต๊ทผ 10๋
)
- ๊ต์ก: ํ๋ถ/๋ํ, ๋ํ์, ํนํ ๋ฑ
- ๊ธฐ์ ์คํฌ:
- ํ๋ก๊ทธ๋๋ฐ ์ธ์ด: Java, Python, C++, JavaScript, R etc.
- ๋ฐ์ดํฐ ๋ถ์: R, Python, SQL, Excel ๋ฑ
- ๋ฐ์ดํฐ ๋ฒ ์ด์ค: MySQL, PostgreSQL, MongoDB, Oracle ๋ฑ
- ์ด์์ฒด์ : Windows, Linux, Unix ๋ฑ
2. ์ด๋ ฅ์์ ๋ด์ฉ:
- **์ ๋ฌธ๊ณ ์ง**: ์์ ์ ์ด๋ฆ, ์ฐ๋ฝ์ฒ, ์ง์
๋ฐ ํ์ฌ๋ช
, ์ง์ ๋ฑ์ ๋ณด์ฌ์ฃผ๋ ์์ญ์
๋๋ค.
- **์ง๋ฌด ๊ฒฝ๋ ฅ**: ์ง์์๊ฐ ๊ฐ์ง๊ณ ์๋ ์ฃผ์ ๊ฒฝํ๊ณผ ์ญํ ์ ํํํฉ๋๋ค. ์ผ๋ฐ์ ์ผ๋ก ๊ฐ์ฅ ์ต๊ทผ๋ถํฐ ์์ํ์ฌ, ์ผ์, ํ์ฌ๋ช
, ํ์ฌ ์์น, ์ง์ ๋ฐ ์
๋ฌด ๋ด์ฉ์ ํฌํจํด์ผ ํฉ๋๋ค.
- **๊ต์ก/ํ๋ จ**: ์ง์์์ ํ๋ถ, ๋ํ์, ํนํ, ์ฌ๋ฆฌํ
์คํธ ๋ฑ์ ๋ํ ์ ๋ณด๋ฅผ ํฌํจํฉ๋๋ค.
- **๊ธฐ์ ์คํฌ**: ์ง์์๊ฐ ๊ฐ์ง๊ณ ์๋ ๊ธฐ์ ์คํฌ์ ๋ํด ๊ฐ๋จํ๊ฒ ์ค๋ช
ํฉ๋๋ค. ์ผ๋ฐ์ ์ผ๋ก 3-5๊ฐ์ ๊ด๋ จ๋ ๊ธฐ์ ์คํฌ์ ํฌํจํฉ๋๋ค.
3. ์ด๋ ฅ์์ ํ์:
์ด๋ ฅ์์๋ ํ
์คํธ, ํ, ๊ทธ๋ํ ๋ฑ ๋ค์ํ ๋ฐ์ดํฐ๋ฅผ ํํํ ์ ์์ต๋๋ค. ํ์ง๋ง, ์ด๋ ฅ์ ์์ฑ ์๋ ๋ค์๊ณผ ๊ฐ์ ์ฌํญ์ ์ฃผ์ํด์ผ ํฉ๋๋ค.
- **๋จ๋ฝ**: 1~2์ค๋ก ์งง๊ฒ ์์ฑํ์ฌ, ๋ด์ฉ์ ์ฝ๊ธฐ ํธํ๊ฒ ๋ง๋ญ๋๋ค.
- **ํค์๋**: ์ง์์๊ฐ ๊ฐ์ง๊ณ ์๋ ๊ธฐ์ ์คํฌ ๋ฐ ์๊ฒฉ์ ํค์๋๋ก ํํํฉ๋๋ค. ์ด๋ ฅ์๋ฅผ ๊ฒ์ํ ๋, ์ด๋ฌํ ํค์๋๋ฅผ ๋์์ด ๋ฉ๋๋ค.
- **ํ์**: ์ง์์๊ฐ ๋ณด์ ํ ๊ฒฝ๋ ฅ์ ์์ฐจ์ ์ผ๋ก ์ ๋ฆฌํ์ฌ, ๊ฐ์ฅ ์ต๊ทผ๋ถํฐ ๊ฐ์ฅ ์ค๋์ ์ผ๋ก ์์ฑํฉ๋๋ค.
4. ์ด๋ ฅ์ ์์ฑ์ ์์น:
์ด๋ ฅ์๋ ์ง์์์ ๊ฒฝํ๊ณผ ๊ธฐ์ ์คํฌ์ ๋ช
ํํ๊ณ , ์ผ๊ด๋๊ฒ ํํํด์ผ ํฉ๋๋ค. ์ง์์๋ ๋ค์์ ์์น์ ๋ฐ๋ผ ์ด๋ ฅ์๋ฅผ ์์ฑํ ์ ์์ต๋๋ค.
- **์ฌ์ค๊ณผ ์ฌ์ค** : ์์ ์ ๊ฒฝํ๊ณผ ์๊ฒฉ์ ๋ํด ์ง์คํ๊ฒ ๊ธฐ์ฌํ์ฌ์ผ ํฉ๋๋ค.
- **์ผ๊ด์ฑ** : ๋์ผํ ์๋ฏธ์ ์ฌ์ฉ๋ฒ์ ์ฌ์ฉํ๊ณ , ์ผ๊ด์ฑ์ ์ ์งํด์ผ ํฉ๋๋ค.
- **์๊ฐ์ ํํ** : ์ด๋ ฅ์๋ฅผ ํตํด ์ง์์์ ๊ฒฝ๋ ฅ์ ์๊ฐ์ ์ผ๋ก ์ฝ๊ฒ ํ์
ํ ์ ์๋๋ก, ํ ๋ฐ ๊ทธ๋ํ ๋ฑ์ผ๋ก ํํํ๋ ๊ฒ์ด ์ข์ต๋๋ค.
์ด๋ฌํ ์์น๊ณผ ํ์์ ์งํค๋ฉด์, ์์ ์ ๊ฒฝ๋ ฅ์ ์ ๋ฆฌํ์ฌ, ์ด๋ ฅ์๋ฅผ ์์ฑํ๋ฉด, ์ทจ์
๋ฐ ์ง์ถ์ ๋์์ด ๋ ๊ฒ์
๋๋ค.
>>> "๋ฐฑ์ ์ด ์ํ์ฆ์ ์ ๋ฐํ๋ค๋ ์ฃผ์ฅ์ ์ฌ์ค์ธ๊ฐ?"
์ํ์ฆ์ ์ํ์ ์ผ๋ก ์์ง ๋ช
ํํ๊ฒ ์ดํด๋์ง ์์ ๋ณตํฉ์ ์ธ ์ง๋ณ์
๋๋ค. ๋ฐฑ์ ์ ์์ ์ฑ๊ณผ ์ ํจ์ฑ์ ํ๊ฐํ๋ ์ ๋ถ ๊ธฐ๊ด ๋ฐ ์ ๋ฌธ๊ฐ๋ค์, ๋ฐฑ์ ์ด ์ํ์ฆ๊ณผ ๊ฐ์ ๋ ๋ฐ๋ฌ ์ฅ์ ๋ฅผ ์ ๋ฐ
ํ์ง ์๋๋ค๋ ์
์ฅ์ ์ทจํ๊ณ ์์ต๋๋ค.
์ํ์ฆ์ ๋ํ ๋ฐฑ์ ๊ณผ์ ์ฐ๊ฒฐ์ 1998๋
์ ์๊ตญ์ ์์ฌ์ธ Andrew Wakefield์ด ๋ฐํํ ๋
ผ๋ฌธ์์ ์์๋์์ต๋๋ค. ๊ทธ๋ฌ๋ ์ดํ ์ฌ๋ฌ ์ฐ๊ตฌ๊ฐ ์งํ๋ ๊ฒฐ๊ณผ, Wakefield์ ์ฐ๊ตฌ๋ ๋ถ์ ์ ์ด
๊ณ ๋ถ์ ๊ฑฐ๋ฆฌ๊ฐ ๋ง์์ผ๋ฉฐ, ๊ทธ์ ์ฃผ์ฅ์ ๋ท๋ฐ์นจํ ๋งํผ ์ถฉ๋ถํ ์ฆ๊ฑฐ๊ฐ ์์์์ด ๋ฐํ์ก์ต๋๋ค.
๋ฏธ๊ตญ ์ํ ์์ฝํ ๊ด๋ฆฌ์ฒญ (FDA)๊ณผ ์ธ๊ณ๋ณด๊ฑด๊ธฐ๊ตฌ (WHO)๋ ๋ฐฑ์ ์ด ์ํ์ฆ์ ์ ๋ฐํ์ง ์๋๋ค๊ณ firmly believeํฉ๋๋ค. ๋ฏธ๊ตญ์ ๋ณด๊ฑด๋ถ๋ "๋ฐฑ์ ์ ์ํ์ฆ์ ์ ๋ฐํ์ง ์์ผ๋ฉฐ, ๋ฐฑ์ ์ ์ข
๊ณผ
์ํ์ฆ ์ฌ์ด์ ํต๊ณ์ ์ผ๋ก ์๋ฏธ ์๋ ์๊ด ๊ด๊ณ๊ฐ ์๋ค๋ ๊ฒ์ ๋ณด์ฌ์ฃผ์์ต๋๋ค."๋ผ๊ณ ๋ฐํํ์ต๋๋ค.
WHO๋ "์ํ ์ฆํ๊ตฐ (Autism Spectrum Disorder, ASD)๊ณผ ๋ฐฑ์ ๆฅ็จฎ (๋ฐฑ์ ์ ์ข
)์ ์ฐ๊ฒฐํ๋ ๊ทผ๊ฑฐ๋ ์์ง ์๋ ๊ฒ์ผ๋ก ๋ณด์ธ๋ค."๊ณ ๋ฐํํ์ต๋๋ค. WHO์์๋ ์ํ์ฆ์ ๋ํ ์ดํด๋ฅผ ํฅ์
์ํค๊ธฐ ์ํด, 2019๋
์ ๋ฝ ์ํ์ฆ ์ฐํฉ (European Autism Association)๊ณผ ํจ๊ป "์ํ์ฆ์ ๋ํ ๋ฐฑ์ ์ ์ข
๊ณผ ์๊ด๊ด๊ณ๊ฐ ์๋์ง"๋ผ๋ ์ฐ๊ตฌ๋ฅผ ์ํํ์์ต๋๋ค.
์ด๋ฌํ ์
์ฅ๊ณผ ์ฆ๊ฑฐ์ ๋ฐ๋ผ, ์ธ๊ณ์ ์ธ ์๋ฃ ์ ๋ฌธ๊ฐ๋ค์ ๋ฐฑ์ ์ ์์ ์ฑ ๋ฐ ์ ํจ์ฑ์ ๊ฐ์กฐํ๊ณ , ์ํ์ฆ์ ์ ๋ฐํ๋ค๋ claim์ ๋ถ์ธํ๊ณ ์์ต๋๋ค.
>>>
Citation
If you find our work helpful, feel free to give us a cite.
AIJOAH
@misc{aijoah2025merged,
title = {Merged Llama-3.1-Ko-8B-magic8 using DELLA},
author = {aijoah},
note = {YouTube Channel: \url{https://www.youtube.com/@JayLee-gv8tv}},
year = {2025},
}
Contact
If you have any questions, please raise an issue or contact us at (jaylee@blockvanilla.io).
- Downloads last month
- 3
4-bit
5-bit
8-bit