Instructions to use prism-ml/Ternary-Bonsai-8B-gguf with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use prism-ml/Ternary-Bonsai-8B-gguf 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 prism-ml/Ternary-Bonsai-8B-gguf:F16
# Run inference directly in the terminal:
llama cli -hf prism-ml/Ternary-Bonsai-8B-gguf:F16

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf prism-ml/Ternary-Bonsai-8B-gguf:F16
# Run inference directly in the terminal:
llama cli -hf prism-ml/Ternary-Bonsai-8B-gguf:F16

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 prism-ml/Ternary-Bonsai-8B-gguf:F16
# Run inference directly in the terminal:
./llama-cli -hf prism-ml/Ternary-Bonsai-8B-gguf:F16

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 prism-ml/Ternary-Bonsai-8B-gguf:F16
# Run inference directly in the terminal:
./build/bin/llama-cli -hf prism-ml/Ternary-Bonsai-8B-gguf:F16

Use Docker

docker model run hf.co/prism-ml/Ternary-Bonsai-8B-gguf:F16

LM Studio
Jan

vLLM

How to use prism-ml/Ternary-Bonsai-8B-gguf with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/prism-ml/Ternary-Bonsai-8B-gguf:F16

Ollama
How to use prism-ml/Ternary-Bonsai-8B-gguf with Ollama:
```
ollama run hf.co/prism-ml/Ternary-Bonsai-8B-gguf:F16
```
Unsloth Desktop

How to use prism-ml/Ternary-Bonsai-8B-gguf with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf prism-ml/Ternary-Bonsai-8B-gguf:F16

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": "prism-ml/Ternary-Bonsai-8B-gguf:F16"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use prism-ml/Ternary-Bonsai-8B-gguf with Docker Model Runner:
```
docker model run hf.co/prism-ml/Ternary-Bonsai-8B-gguf:F16
```

Lemonade

How to use prism-ml/Ternary-Bonsai-8B-gguf with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull prism-ml/Ternary-Bonsai-8B-gguf:F16

Run and chat with the model

lemonade run user.Ternary-Bonsai-8B-gguf-F16

List all available models

lemonade list

Hermes Agent

How to use prism-ml/Ternary-Bonsai-8B-gguf with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf prism-ml/Ternary-Bonsai-8B-gguf:F16

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 prism-ml/Ternary-Bonsai-8B-gguf:F16

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use prism-ml/Ternary-Bonsai-8B-gguf with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf prism-ml/Ternary-Bonsai-8B-gguf:F16

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 "prism-ml/Ternary-Bonsai-8B-gguf:F16" \
  --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"

The reported accuracy could not be replicated.

by DiyanGoodLuck - opened Jun 11

Discussion

DiyanGoodLuck

Jun 11

I tried to replicate the results for GSM8K, IFEval, Humaneval+, and MBPP using default lm-eval config, but the numbers I got are way off from what's in the white paper. Is there a specific setup I can use to get results that match the white paper?

This is the command I use to start llama-cpp server:

    nohup ${LLAMA_CPP_DIR}/build/bin/llama-server \
    -m ${MODEL_PATH} \
    --port ${SERVER_PORT} \
    -c 4096 \
    --log-disable \
    > ${SERVER_LOG} 2>&1 &

IFEval, Humaneval+, and MBPP tasks are similar, but I did not set num_fewshot for them.

model: gguf
model_args:
  base_url: http://localhost:8080
tasks:
  - gsm8k
num_fewshot: 5
batch_size: 1

Tasks	Version	Filter	n-shot	Metric		Value		Stderr
humaneval_plus	1	create_test	0	pass@1	↑	0.2195	±	0.0324
mbpp_plus	1	none	0	pass_at_1	↑	0.0000	±	0.0000
gsm8k	3	flexible-extract	5	exact_match	↑	0.8491	±	0.0099
		strict-match	5	exact_match	↑	0.8506	±	0.0098
ifeval	4	none	0	inst_level_loose_acc	↑	0.4472	±	N/A
		none	0	inst_level_strict_acc	↑	0.4089	±	N/A
		none	0	prompt_level_loose_acc	↑	0.3216	±	0.0201
		none	0	prompt_level_strict_acc	↑	0.2717	±	0.0191

pashak

Prism ML org Jun 12

You can check the appendix (B. Benchmark Evaluation Methodology) for more details on the eval setup:
All the eval numbers in the paper were ran through the same setup.
https://github.com/PrismML-Eng/Bonsai-demo/blob/main/1-bit-bonsai-8b-whitepaper.pdf

KirMas

Jul 8

@diyangoodluck

Can you share your findings on the performance?

I was doing more prompt-based comparison to other similar-sized models in LM Studio and Bonsai-8B was lagging behind. I am not sure if the reason in backend of llama.cpp or there is Bonsai performance issue, so I'm curious what was your experience.

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