Instructions to use wish418/carbix-v6 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- llama.cpp
How to use wish418/carbix-v6 with llama.cpp:
Install (macOS, Linux)
curl -LsSf https://llama.app/install.sh | sh # Start a local OpenAI-compatible server with a web UI: llama serve -hf wish418/carbix-v6 # Run inference directly in the terminal: llama cli -hf wish418/carbix-v6
Install from WinGet (Windows)
winget install llama.cpp # Start a local OpenAI-compatible server with a web UI: llama serve -hf wish418/carbix-v6 # Run inference directly in the terminal: llama cli -hf wish418/carbix-v6
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 wish418/carbix-v6 # Run inference directly in the terminal: ./llama-cli -hf wish418/carbix-v6
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 wish418/carbix-v6 # Run inference directly in the terminal: ./build/bin/llama-cli -hf wish418/carbix-v6
Use Docker
docker model run hf.co/wish418/carbix-v6
- LM Studio
- Jan
- Ollama
How to use wish418/carbix-v6 with Ollama:
ollama run hf.co/wish418/carbix-v6
- Unsloth Desktop
- Pi
How to use wish418/carbix-v6 with Pi:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama serve -hf wish418/carbix-v6
Configure the model in Pi
# Install Pi: npm install -g @earendil-works/pi-coding-agent # Add to ~/.pi/agent/models.json: { "providers": { "llama-cpp": { "baseUrl": "http://localhost:8080/v1", "api": "openai-completions", "apiKey": "none", "models": [ { "id": "wish418/carbix-v6" } ] } } }Run Pi
# Start Pi in your project directory: pi
- Docker Model Runner
How to use wish418/carbix-v6 with Docker Model Runner:
docker model run hf.co/wish418/carbix-v6
- Lemonade
How to use wish418/carbix-v6 with Lemonade:
Pull the model
# Download Lemonade from https://lemonade-server.ai/ lemonade pull wish418/carbix-v6
Run and chat with the model
lemonade run user.carbix-v6-{{QUANT_TAG}}List all available models
lemonade list
- Hermes Agent
How to use wish418/carbix-v6 with Hermes Agent:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama serve -hf wish418/carbix-v6
Configure Hermes
# Install Hermes: curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash hermes setup # Point Hermes at the local server: hermes config set model.provider custom hermes config set model.base_url http://127.0.0.1:8080/v1 hermes config set model.default wish418/carbix-v6
Run Hermes
hermes
- Atomic Chat
- OpenClaw
How to use wish418/carbix-v6 with OpenClaw:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama serve -hf wish418/carbix-v6
Configure OpenClaw
# Install OpenClaw: npm install -g openclaw@latest # Register the local server and set it as the default model: openclaw onboard --non-interactive --mode local \ --auth-choice custom-api-key \ --custom-base-url http://127.0.0.1:8080/v1 \ --custom-model-id "wish418/carbix-v6" \ --custom-provider-id llama-cpp \ --custom-compatibility openai \ --custom-text-input \ --accept-risk \ --skip-health
Run OpenClaw
openclaw agent --local --agent main --message "Hello from Hugging Face"
carbix-v6
ํ๊ตญ ํ์๋ฐฐ์ถ ์ฆ๋น ๋ฌธ์(์ ๊ธฐ์๊ธ ์ฒญ๊ตฌ์ ยท ๋์๊ฐ์ค ์ฒญ๊ตฌ์ ยท ํ๋ฌผ ์ด์ก์ฆ)์์ ๊ตฌ์กฐํ JSON์ ์ถ์ถํ๋ Qwen3-VL-4B LoRA ํ์ธํ๋ ๋ชจ๋ธ.
v6๋ ์นดํ ๊ณ ๋ฆฌ ํํธ ์์ด ๋ฌธ์ ์ข ๋ฅ๋ฅผ ์ค์ค๋ก ํ๋ณํ๋ฉด์ ์ถ์ถ ์ ํ๋๋ฅผ ์ ์งํ๋ค.
๊ตฌ์ฑ
| ํ์ผ | ํฌ๊ธฐ | ์ค๋ช |
|---|---|---|
carbix-v6-q8.gguf |
4.28GB | llama.cpp ์๋น์ฉ ๊ฐ์ค์น (q8_0) |
carbix-v6-mmproj.gguf |
836MB | ๋น์ ํ๋ก์ ํฐ (f16) |
lora_adapter_v6/ |
132MB | PEFT ์ด๋ํฐ (merge ์ ) |
v5 ๋๋น ๋ณ๊ฒฝ
- ํ์ต ํ๋กฌํํธ๋ฅผ ์นดํ
๊ณ ๋ฆฌ ํํธ ์๋ ์๋๋ถ๋ฅ์ฉ 1์ข
์ผ๋ก ํต์ผ. v5๊น์ง๋
์นดํ
๊ณ ๋ฆฌ๋ณ 3์ข
(์ ๊ธฐ 2,167์ ยท ๊ฐ์ค 2,677์ ยท ์ด์ก 4,782์)์ ์ผ๋๋ฐ, ๊ทธ๋ฌ๋ฉด
์ ๋ต์
category๊ฐ ํ๋กฌํํธ์ ์ด๋ฏธ ์ ํ ๊ฐ์ ๋ณต์ฌ๋ผ ๋ชจ๋ธ์ด ๋ถ๋ฅ๋ฅผ ๋ฐฐ์ธ ๊ธฐํ๊ฐ ์์๋ค. v6๋ ์ ์ํ์ด ๋์ผํ 5,241์ ์๋๋ถ๋ฅ ํ๋กฌํํธ๋ค. - ํ๋กฌํํธ v13 (์คํค๋ง ๋ณ๊ฒฝ ์์)
ํ์ต
๋ฒ ์ด์ค Qwen/Qwen3-VL-4B-Instruct
LoRA r=16, alpha=32, dropout=0.05
q/k/v/o_proj, gate/up/down_proj
์ต์ ํ AdamW, LR 2e-4, cosine, warmup 0.05 (246์คํ
)
๋ฐฐ์น 1 ร grad_accum 4 (effective 4)
์คํ
4,925 (1 epoch), 15์๊ฐ 41๋ถ (L4 24GB)
๋ฐ์ดํฐ 19,700๊ฑด โ ๊ฐ์ค 10,727 ยท ์ ๊ธฐ 4,495 ยท ์ด์ก 4,478
์์ค train 0.0087 / eval 0.0020
๋ถ๋ฅ ์ ํ๋ (ํํธ ์๋ ํ๋กฌํํธ, 120๊ฑด)
| ๋ชจ๋ธ | ์ ํ๋ | ์ค๋ถ๋ฅ |
|---|---|---|
| Qwen3-VL-4B (ํ์ธํ๋ ์ ) | 87.5% | 15๊ฑด |
| carbix-v5 | 56.7% | 52๊ฑด |
| carbix-v6 | 100.0% | 0๊ฑด |
์ ๊ธฐ 9/9 ยท ๊ฐ์ค 75/75 ยท ์ด์ก 36/36. unknown ์ถ๋ ฅ, ํ ํฐ ์ด๊ณผ, ํ์ฑ ์คํจ ๋ชจ๋ 0๊ฑด.
v5๊ฐ ๋ฒ ์ด์ค๋ณด๋ค ๋ฎ์ ๊ฒ์ด ํํธ ์ ์ฉ ํ์ต์ ๋๊ฐ์๋ค โ ํ๋กฌํํธ๊ฐ ํญ์ ์ ๋ต์ ์๋ ค์คฌ์ผ๋ฏ๋ก ๋ถ๋ฅ๋ฅผ ์ํํ ํ์๊ฐ ์์๊ณ , ํํธ ์๋ ์ ๋ ฅ์ ํ์ต ๋ถํฌ ๋ฐ์ด์๋ค.
์ถ์ถ ์ ํ๋ (ํ ์คํธ 657๊ฑด, llama.cpp q8_0)
| ์นดํ ๊ณ ๋ฆฌ | v4 | v5 | v6 (์๋๋ถ๋ฅ) |
|---|---|---|---|
| ์ ๊ธฐ | 100.0% | 100.0% | 100.0% |
| ๊ฐ์ค | 99.85% | 99.87% | 99.85% |
| ์ด์ก | 95.43% | 98.00% | 98.14% |
| ์ ์ฒด | 98.76% | 99.42% | 99.43% |
| ANLS | 0.9954 | 0.9979 | 0.9984 |
| Token-F1 | 0.9898 | 0.9947 | 0.9946 |
v6 ์ด์ ์นดํ ๊ณ ๋ฆฌ๋ฅผ ์ฌ์ ์ง์ ํ์ง ์๊ณ ์ธก์ ํ ๊ฐ์ด๋ค. v4ยทv5 ๋ ์ ๋ต ์นดํ ๊ณ ๋ฆฌ๋ฅผ ํ๋กฌํํธ๋ก ๋ฐ์ ์ํ์ ์์น์ด๋ค. ์ค๋ถ๋ฅ๊ฐ ์์๋ค๋ฉด ํด๋น ๋ฌธ์์ ํ๋๊ฐ ํต์งธ๋ก ์ด๊ธ๋ ์ ์๊ฐ ์ ์ง๋ ์ ์๋ค
์นดํ ๊ณ ๋ฆฌ ํํธ๋ฅผ ์ฃผ๋ ๊ธฐ์กด ๊ฒฝ๋ก๋ ์ ์ง๋๋ค: 807๊ฑด ๊ธฐ์ค 98.82%(ANLS 0.9962)๋ก v5์ 98.87% ์ ์ฌ์ค์ ๋์ผํ๋ค.
๊ฑด๋น ํ๊ท 5.0์ด (p50 3.5s / p95 11.0s).
์๋น
llama-server -m carbix-v6-q8.gguf --mmproj carbix-v6-mmproj.gguf \
--host 127.0.0.1 --port 8001 -ngl 99 -c 8192 \
--parallel 2 --image-min-tokens 1024 --jinja
max_tokens ๋ 768 ์ด์์ ๊ถ์ฅ โ ์ด์ก 25ํค ์ถ๋ ฅ์ด ๊ฒฝ์ ์ง 3๊ฐ์ผ ๋
461 ํ ํฐ๊น์ง ๋ฐ์
- Downloads last month
- 119
We're not able to determine the quantization variants.
Model tree for wish418/carbix-v6
Base model
Qwen/Qwen3-VL-4B-Instruct