Instructions to use meditsolutions/Llama-3.2-SUN-1B-Instruct with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use meditsolutions/Llama-3.2-SUN-1B-Instruct with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="meditsolutions/Llama-3.2-SUN-1B-Instruct")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("meditsolutions/Llama-3.2-SUN-1B-Instruct")
model = AutoModelForCausalLM.from_pretrained("meditsolutions/Llama-3.2-SUN-1B-Instruct", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use meditsolutions/Llama-3.2-SUN-1B-Instruct 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 meditsolutions/Llama-3.2-SUN-1B-Instruct:Q8_0
# Run inference directly in the terminal:
llama cli -hf meditsolutions/Llama-3.2-SUN-1B-Instruct:Q8_0

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf meditsolutions/Llama-3.2-SUN-1B-Instruct:Q8_0
# Run inference directly in the terminal:
llama cli -hf meditsolutions/Llama-3.2-SUN-1B-Instruct:Q8_0

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 meditsolutions/Llama-3.2-SUN-1B-Instruct:Q8_0
# Run inference directly in the terminal:
./llama-cli -hf meditsolutions/Llama-3.2-SUN-1B-Instruct:Q8_0

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 meditsolutions/Llama-3.2-SUN-1B-Instruct:Q8_0
# Run inference directly in the terminal:
./build/bin/llama-cli -hf meditsolutions/Llama-3.2-SUN-1B-Instruct:Q8_0

Use Docker

docker model run hf.co/meditsolutions/Llama-3.2-SUN-1B-Instruct:Q8_0

LM Studio
Jan

vLLM

How to use meditsolutions/Llama-3.2-SUN-1B-Instruct with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "meditsolutions/Llama-3.2-SUN-1B-Instruct"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "meditsolutions/Llama-3.2-SUN-1B-Instruct",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/meditsolutions/Llama-3.2-SUN-1B-Instruct:Q8_0

SGLang

How to use meditsolutions/Llama-3.2-SUN-1B-Instruct with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "meditsolutions/Llama-3.2-SUN-1B-Instruct" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "meditsolutions/Llama-3.2-SUN-1B-Instruct",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "meditsolutions/Llama-3.2-SUN-1B-Instruct" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "meditsolutions/Llama-3.2-SUN-1B-Instruct",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Ollama
How to use meditsolutions/Llama-3.2-SUN-1B-Instruct with Ollama:
```
ollama run hf.co/meditsolutions/Llama-3.2-SUN-1B-Instruct:Q8_0
```
Unsloth Desktop

How to use meditsolutions/Llama-3.2-SUN-1B-Instruct with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf meditsolutions/Llama-3.2-SUN-1B-Instruct:Q8_0

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": "meditsolutions/Llama-3.2-SUN-1B-Instruct:Q8_0"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use meditsolutions/Llama-3.2-SUN-1B-Instruct with Docker Model Runner:
```
docker model run hf.co/meditsolutions/Llama-3.2-SUN-1B-Instruct:Q8_0
```

Lemonade

How to use meditsolutions/Llama-3.2-SUN-1B-Instruct with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull meditsolutions/Llama-3.2-SUN-1B-Instruct:Q8_0

Run and chat with the model

lemonade run user.Llama-3.2-SUN-1B-Instruct-Q8_0

List all available models

lemonade list

Hermes Agent

How to use meditsolutions/Llama-3.2-SUN-1B-Instruct with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf meditsolutions/Llama-3.2-SUN-1B-Instruct:Q8_0

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 meditsolutions/Llama-3.2-SUN-1B-Instruct:Q8_0

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use meditsolutions/Llama-3.2-SUN-1B-Instruct with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf meditsolutions/Llama-3.2-SUN-1B-Instruct:Q8_0

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 "meditsolutions/Llama-3.2-SUN-1B-Instruct:Q8_0" \
  --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"

MedIT SUN 1B Instruct

Base Model

Llama 3.2 1B -> MedIT SUN 2.5B -> MedIT SUN 1B -> Knowledge Injection from Llama 3.1 8B Instruct

Mesh Size

1B to 2.5B parameters MedIT SUN 2.5B -> layers mesh using MedIT-mesh technique and downscaled to 1B

Extension Method

Proprietary technique developed by MedIT Solutions

Fine-tuning

Open (or open subsets allowing for commercial use) open datasets from HF
Open (or open subsets allowing for commercial use) SFT datasets from HF

Training Status

Current version: instruct-1.0.0

Key Features

Built on Llama 3.2 architecture
Upscaled from 1B to 2.47B parameters
Optimized for open-ended conversations
Incorporates supervised fine-tuning for improved performance
Layers meshing using the MedIT-mesh technique
Downscaled to 1B
Knowledge injection from Llama 3.1 8B Instruct using new technique developed by MedIT Solutions

Use Case

General conversation and task-oriented interactions

Limitations As the model is still in training, performance and capabilities may vary. Users should be aware that the model is not in its final form and may exhibit inconsistencies or limitations typical of in-progress AI models.

Disclaimer and Safety Considerations The Model is designed to be used as a smart assistant but not as a knowledge source within your applications, systems, or environments. It is not intended to provide 100% accurate answers, especially in scenarios where high precision and accuracy are

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	15.11
IFEval (0-Shot)	64.13
BBH (3-Shot)	9.18
MATH Lvl 5 (4-Shot)	4.61
GPQA (0-shot)	0.00
MuSR (0-shot)	4.05
MMLU-PRO (5-shot)	8.68

Downloads last month: 39

Safetensors

Model size

1B params

Tensor type

BF16

Model tree for meditsolutions/Llama-3.2-SUN-1B-Instruct

Base model

meta-llama/Llama-3.2-1B-Instruct

Quantized

(422)

this model

Datasets used to train meditsolutions/Llama-3.2-SUN-1B-Instruct

Collection including meditsolutions/Llama-3.2-SUN-1B-Instruct

MedIT SUN

Collection

Llama 3.2 1B upscaled to 2.5B parameters • 4 items • Updated Nov 27, 2024 • 1

Evaluation results

strict accuracy on IFEval (0-Shot)
Open LLM Leaderboard

64.130
normalized accuracy on BBH (3-Shot)
Open LLM Leaderboard

9.180
exact match on MATH Lvl 5 (4-Shot)
Open LLM Leaderboard

4.610
acc_norm on GPQA (0-shot)
Open LLM Leaderboard

0.000
acc_norm on MuSR (0-shot)
Open LLM Leaderboard

4.050
accuracy on MMLU-PRO (5-shot)
test set Open LLM Leaderboard

8.680