Instructions to use vinpix/Bonsai-8B-llama.cpp with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use vinpix/Bonsai-8B-llama.cpp 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 vinpix/Bonsai-8B-llama.cpp:Q2_K
# Run inference directly in the terminal:
llama cli -hf vinpix/Bonsai-8B-llama.cpp:Q2_K

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf vinpix/Bonsai-8B-llama.cpp:Q2_K
# Run inference directly in the terminal:
llama cli -hf vinpix/Bonsai-8B-llama.cpp:Q2_K

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 vinpix/Bonsai-8B-llama.cpp:Q2_K
# Run inference directly in the terminal:
./build/bin/llama cli -hf vinpix/Bonsai-8B-llama.cpp:Q2_K

Use Docker

docker model run hf.co/vinpix/Bonsai-8B-llama.cpp:Q2_K

Llama new
LM Studio
Jan

vLLM

How to use vinpix/Bonsai-8B-llama.cpp with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "vinpix/Bonsai-8B-llama.cpp"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "vinpix/Bonsai-8B-llama.cpp",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/vinpix/Bonsai-8B-llama.cpp:Q2_K

Ollama
How to use vinpix/Bonsai-8B-llama.cpp with Ollama:
```
ollama run hf.co/vinpix/Bonsai-8B-llama.cpp:Q2_K
```
Unsloth Desktop

How to use vinpix/Bonsai-8B-llama.cpp with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf vinpix/Bonsai-8B-llama.cpp:Q2_K

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": "vinpix/Bonsai-8B-llama.cpp:Q2_K"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use vinpix/Bonsai-8B-llama.cpp with Docker Model Runner:
```
docker model run hf.co/vinpix/Bonsai-8B-llama.cpp:Q2_K
```

Lemonade

How to use vinpix/Bonsai-8B-llama.cpp with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull vinpix/Bonsai-8B-llama.cpp:Q2_K

Run and chat with the model

lemonade run user.Bonsai-8B-llama.cpp-Q2_K

List all available models

lemonade list

Hermes Agent

How to use vinpix/Bonsai-8B-llama.cpp with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf vinpix/Bonsai-8B-llama.cpp:Q2_K

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 vinpix/Bonsai-8B-llama.cpp:Q2_K

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use vinpix/Bonsai-8B-llama.cpp with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf vinpix/Bonsai-8B-llama.cpp:Q2_K

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 "vinpix/Bonsai-8B-llama.cpp:Q2_K" \
  --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"

Update note on ik_llama.cpp: Lossless repack to PQ2_0 enables native CPU (-rtr) and CUDA offload

by Gaby8999 - opened about 4 hours ago

Discussion

Gaby8999

about 4 hours ago

•

edited about 4 hours ago

Hi! First off, thank you so much for this repo, the Q2_KT build, and the brilliant mathematical explanation of the ternary zero-shift issue. Your analysis was spot-on.

When you released this build back in July, you were 100% right: TQ2_0 couldn't run on ik_llama.cpp, and native ternary was strictly stuck on CPU.

I just wanted to share some recent good news with you: the ecosystem has caught up over the past few weeks, partly inspired by work like yours!

Lossless repack script (TQ2_0 ➔ PQ2_0):
Mainline uses $QK=256$ for TQ2_0 while ik_llama uses $QK=128$ for Prism's PQ2_0. I recently contributed scripts/repack_ternary_to_prism.py (merged in PR #2694), which losslessly converts your TQ2_0 GGUF to PQ2_0 in seconds (exact trits preserved, scales duplicated).
Native CUDA kernels landed:
In recent commits (PR #2691 & #2747), ik_llama.cpp finally added native CUDA support for PQ2_0. So repacked native ternary can now actually be offloaded to Nvidia GPUs without token degradation.
CPU speed:
On CPU, running the repacked file with -rtr triggers the interleaved PQ2_0_R8 path, giving a huge boost to prompt processing.

No obligation or pressure to change anything on your end! I just wanted to share this so you know your work helped push things forward. If you ever feel like updating the note or offering a PQ2_0 build for ik_llama users down the road, the tools are officially in main now.

Thanks again for the awesome work on Bonsai!

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment