How to use from
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 ZichenAI/BaiHu-gguf:F16
# Run inference directly in the terminal:
llama cli -hf ZichenAI/BaiHu-gguf:F16
Install from WinGet (Windows)
winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf ZichenAI/BaiHu-gguf:F16
# Run inference directly in the terminal:
llama cli -hf ZichenAI/BaiHu-gguf:F16
Build from source code
git clone https://github.com/ggml-org/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-app
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama serve -hf ZichenAI/BaiHu-gguf:F16
# Run inference directly in the terminal:
./build/bin/llama cli -hf ZichenAI/BaiHu-gguf:F16
Use Docker
docker model run hf.co/ZichenAI/BaiHu-gguf:F16
Quick Links

็™ฝ่™Ž-v2

็™ฝ่™Ž-v2 ๆ˜ฏไธ€ๆฌพๅคšๆจกๆ€ๅคง่ฏญ่จ€ๆจกๅž‹๏ผŒๆ”ฏๆŒๆ–‡ๆœฌใ€ๅ›พๅƒใ€้Ÿณ้ข‘ไธŽ่ง†้ข‘่พ“ๅ…ฅ๏ผŒ้€‚็”จไบŽไธญๆ–‡ๅœบๆ™ฏไธ‹็š„ๅคšๆจกๆ€็†่งฃไธŽ็”ŸๆˆไปปๅŠกใ€‚


ๆจกๅž‹็ฎ€ไป‹

  • ๆจกๅž‹ๅ็งฐ: ็™ฝ่™Ž-v2
  • ไธŠไธ‹ๆ–‡้•ฟๅบฆ: 131,072 tokens
  • ่ฏ่กจๅคงๅฐ: 262,144
  • ๆ•ฐๆฎ็ฑปๅž‹: float16
  • ๆ”ฏๆŒๆจกๆ€: ๆ–‡ๆœฌ / ๅ›พๅƒ / ้Ÿณ้ข‘ / ่ง†้ข‘

ไป“ๅบ“ๆ–‡ไปถ่ฏดๆ˜Ž

ๆ–‡ไปถ ่ฏดๆ˜Ž
BaiHu-v2.Q4_K_M.gguf ไธปๆจกๅž‹ GGUF ้‡ๅŒ–็‰ˆๆœฌ๏ผˆQ4_K_M๏ผ‰๏ผŒ้€‚ๅˆๆœฌๅœฐ CPU/GPU ๆŽจ็†
BaiHu-v2.F16-mmproj.gguf ๅคšๆจกๆ€ๆŠ•ๅฝฑๅฑ‚๏ผˆmmproj๏ผ‰FP16 ็‰ˆๆœฌ๏ผŒ้…ๅˆไธปๆจกๅž‹็”จไบŽๅ›พๅƒ/้Ÿณ้ข‘/่ง†้ข‘็†่งฃ
Modelfile llama.cpp / Ollama ็š„ๆจกๅž‹้…็ฝฎๆ–‡ไปถ็คบไพ‹

ๆŽจ่ๆญ้…ไฝฟ็”จ๏ผšBaiHu-v2.Q4_K_M.gguf + BaiHu-v2.F16-mmproj.gguf


ๆจกๅž‹่ƒฝๅŠ›

  • ไธญๆ–‡ๅคš่ฝฎๅฏน่ฏ
  • ๅ›พๅƒๆ่ฟฐไธŽ่ง†่ง‰้—ฎ็ญ”
  • ้Ÿณ้ข‘ๅ†…ๅฎน็†่งฃ
  • ่ง†้ข‘ๅ†…ๅฎน็†่งฃ
  • ๅทฅๅ…ท่ฐƒ็”จ

ไฝฟ็”จๆ–นๆณ•

ไฝฟ็”จ llama.cpp / Ollama ๆŽจ็†

ๅ‚่€ƒไป“ๅบ“ไธญ็š„ Modelfile ๅˆ›ๅปบ Ollama ๆจกๅž‹๏ผš

ollama create BaiHu-v2 -f Modelfile
ollama run BaiHu-v2

ไฝฟ็”จ llama.cpp ๅ‘ฝไปค่กŒ

./llama-cli \
  -m BaiHu-v2.Q4_K_M.gguf \
  --mmproj BaiHu-v2.F16-mmproj.gguf \
  --image example.jpg \
  -p "่ฏทๆ่ฟฐ่ฟ™ๅผ ๅ›พ็‰‡๏ผš"

ไฝฟ็”จ transformers๏ผˆๅฎŒๆ•ดๆจกๅž‹๏ผ‰

ๅฎŒๆ•ด PyTorch/Safetensors ็‰ˆๆœฌ่ฏทๅ‚่€ƒ้…ๅฅ—ไป“ๅบ“ใ€‚ๆœฌไป“ๅบ“ไป…ๆไพ› GGUF ้‡ๅŒ–็‰ˆๆœฌใ€‚


ๆจกๅž‹้…็ฝฎ

  • ๆ–‡ๆœฌๆจกๅž‹: 35 ๅฑ‚๏ผŒ้š่—็ปดๅบฆ 1536๏ผŒ8 ๅคดๆณจๆ„ๅŠ›
  • ่ง†่ง‰็ผ–็ ๅ™จ: 16 ๅฑ‚๏ผŒ้š่—็ปดๅบฆ 768๏ผŒๅ›พๅƒ token ๆ•ฐ 280
  • ้Ÿณ้ข‘็ผ–็ ๅ™จ: 12 ๅฑ‚๏ผŒ้š่—็ปดๅบฆ 1024
  • ่ง†้ข‘: ๆ”ฏๆŒ 32 ๅธง้‡‡ๆ ท๏ผŒๆฏๅธงๆœ€ๅคง 70 ไธช soft token

่ฎญ็ปƒไฟกๆฏ

  • ่ฎญ็ปƒๆก†ๆžถ: Unsloth
  • Unsloth ็‰ˆๆœฌ: 2026.6.8
  • ไผ˜ๅŒ–็›ฎๆ ‡: ๅœจไฟๆŒๅคšๆจกๆ€่ƒฝๅŠ›็š„ๅŒๆ—ถ๏ผŒๆๅ‡ไธญๆ–‡ๆŒ‡ไปค่ทŸ้šไธŽๅฏน่ฏ่ดจ้‡

ๅ…่ดฃๅฃฐๆ˜Ž

ๆœฌๆจกๅž‹็”Ÿๆˆ็š„ๅ†…ๅฎนๅฏ่ƒฝๅ—่ฎญ็ปƒๆ•ฐๆฎๅฝฑๅ“ใ€‚่ฏทๅ‹ฟๅฐ†ๆจกๅž‹่พ“ๅ‡บไฝœไธบไธ“ไธšๅปบ่ฎฎ๏ผˆๅŒป็–—ใ€ๆณ•ๅพ‹ใ€้‡‘่ž็ญ‰๏ผ‰ไฝฟ็”จใ€‚ๆจกๅž‹ๅฏ่ƒฝๅญ˜ๅœจๅนป่ง‰ใ€ๅ่งๆˆ–ไธๅ‡†็กฎไฟกๆฏ๏ผŒ่ฏท่ฐจๆ…Žไฝฟ็”จๅนถ่‡ช่กŒ้ชŒ่ฏใ€‚


ๆŽˆๆƒๅ่ฎฎ

ๆœฌๆจกๅž‹้‡‡็”จ MIT ๅ่ฎฎ ๅผ€ๆบใ€‚ไฝฟ็”จๆœฌๆจกๅž‹ๅ‰่ฏทไป”็ป†้˜…่ฏปๅนถ้ตๅฎˆ MIT ่ฎธๅฏๅ่ฎฎๆกๆฌพใ€‚


่‡ด่ฐข

Downloads last month
151
GGUF
Model size
5B params
Architecture
gemma4
Hardware compatibility
Log In to add your hardware

4-bit

Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support