Instructions to use jxx123/loop-qwen-v8 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use jxx123/loop-qwen-v8 with Transformers:

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

pipe = pipeline("text-generation", model="jxx123/loop-qwen-v8")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("jxx123/loop-qwen-v8")
model = AutoModelForCausalLM.from_pretrained("jxx123/loop-qwen-v8", 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 jxx123/loop-qwen-v8 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 jxx123/loop-qwen-v8:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf jxx123/loop-qwen-v8:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf jxx123/loop-qwen-v8:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf jxx123/loop-qwen-v8:Q4_K_M

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 jxx123/loop-qwen-v8:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf jxx123/loop-qwen-v8:Q4_K_M

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 jxx123/loop-qwen-v8:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf jxx123/loop-qwen-v8:Q4_K_M

Use Docker

docker model run hf.co/jxx123/loop-qwen-v8:Q4_K_M

LM Studio
Jan

vLLM

How to use jxx123/loop-qwen-v8 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "jxx123/loop-qwen-v8"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "jxx123/loop-qwen-v8",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/jxx123/loop-qwen-v8:Q4_K_M

SGLang

How to use jxx123/loop-qwen-v8 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 "jxx123/loop-qwen-v8" \
    --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": "jxx123/loop-qwen-v8",
		"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 "jxx123/loop-qwen-v8" \
        --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": "jxx123/loop-qwen-v8",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Ollama
How to use jxx123/loop-qwen-v8 with Ollama:
```
ollama run hf.co/jxx123/loop-qwen-v8:Q4_K_M
```

Unsloth Studio

How to use jxx123/loop-qwen-v8 with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for jxx123/loop-qwen-v8 to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for jxx123/loop-qwen-v8 to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for jxx123/loop-qwen-v8 to start chatting

How to use jxx123/loop-qwen-v8 with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf jxx123/loop-qwen-v8:Q4_K_M

Configure the model in Pi

# Install Pi:
npm install -g @mariozechner/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "jxx123/loop-qwen-v8:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use jxx123/loop-qwen-v8 with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf jxx123/loop-qwen-v8:Q4_K_M

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 jxx123/loop-qwen-v8:Q4_K_M

Run Hermes

hermes

OpenClaw new

How to use jxx123/loop-qwen-v8 with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf jxx123/loop-qwen-v8:Q4_K_M

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 "jxx123/loop-qwen-v8:Q4_K_M" \
  --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"

Docker Model Runner
How to use jxx123/loop-qwen-v8 with Docker Model Runner:
```
docker model run hf.co/jxx123/loop-qwen-v8:Q4_K_M
```

Lemonade

How to use jxx123/loop-qwen-v8 with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull jxx123/loop-qwen-v8:Q4_K_M

Run and chat with the model

lemonade run user.loop-qwen-v8-Q4_K_M

List all available models

lemonade list

Atomic Chat

You need to agree to share your contact information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

loop-qwen-v8 — on-device insulin-dosing controller (Qwen3-4B)

A Qwen3-4B policy distilled from a Gemini-3-flash-preview insulin-control teacher, for closed-loop Type-1-diabetes insulin dosing. It runs on-device (2.5 GB Q4 GGUF via Ollama/llama.cpp; one decision per 5-min loop) and reaches teacher parity in the simglucose closed-loop simulator.

Results (held-out 9 patients, 48 h, seed 42)

	TIR (70–180)	Survival	TBR (<70)
Gemini teacher	73.5%	8/9	7.0%
loop-qwen-v8 (this model)	72.4%	9/9	9.5%

Confirmed on a novel meal scenario (seed 99): 73.0% TIR, 9/9 survival. All-30 patients: v8 75.8% TIR / 28-30 survival vs Gemini 76.6% / 29-30.

Files

model.safetensors + config/tokenizer — merged fp16 (for transformers / re-quantizing)
loop-qwen-v8-Q4_K_M.gguf — deployable 4-bit (Ollama / llama.cpp)

Usage (Ollama)

ollama create loop-qwen-v8 --quantize q4_K_M -f Modelfile   # or import the gguf
# prompt = patient metadata + 6h CGM/insulin/carb history (JSON)
# output = {"reasoning": "...", "basal_chunk":[5], "bolus_chunk":[5]}

Deployment applies safety clamps: basal ∈ [0,5] U/hr, bolus ∈ [0,20] U.

Recipe

Qwen3-4B + LoRA (r=32) SFT on Gemini Chain-of-Draft demonstrations, +DAgger (relabel student-visited states), + multi-seed meal diversity (the key to matching the teacher without dosing oscillation). Full log: docs/distill_experiment_log.md in the loop-gpt repo.

⚠️ Not a medical device

Research artifact evaluated only in simulation. Residual hypoglycemia (TBR ~10% vs teacher 7%) remains; a low-glucose-suspend safety rail is recommended before any real use. Do not use for actual insulin dosing.

Downloads last month: 14

Safetensors

Model size

4B params

Tensor type

F16

Model tree for jxx123/loop-qwen-v8

Base model

Qwen/Qwen3-4B-Base

Finetuned

Qwen/Qwen3-4B

Quantized

(283)

this model