Instructions to use asharrm/juke-1.7b-dev-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 asharrm/juke-1.7b-dev-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 asharrm/juke-1.7b-dev-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf asharrm/juke-1.7b-dev-GGUF:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf asharrm/juke-1.7b-dev-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf asharrm/juke-1.7b-dev-GGUF:Q4_K_M

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 asharrm/juke-1.7b-dev-GGUF:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama cli -hf asharrm/juke-1.7b-dev-GGUF:Q4_K_M

Use Docker

docker model run hf.co/asharrm/juke-1.7b-dev-GGUF:Q4_K_M

Llama new
LM Studio
Jan

vLLM

How to use asharrm/juke-1.7b-dev-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "asharrm/juke-1.7b-dev-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": "asharrm/juke-1.7b-dev-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/asharrm/juke-1.7b-dev-GGUF:Q4_K_M

Ollama
How to use asharrm/juke-1.7b-dev-GGUF with Ollama:
```
ollama run hf.co/asharrm/juke-1.7b-dev-GGUF:Q4_K_M
```
Unsloth Desktop

How to use asharrm/juke-1.7b-dev-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf asharrm/juke-1.7b-dev-GGUF:Q4_K_M

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": "asharrm/juke-1.7b-dev-GGUF:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use asharrm/juke-1.7b-dev-GGUF with Docker Model Runner:
```
docker model run hf.co/asharrm/juke-1.7b-dev-GGUF:Q4_K_M
```

Lemonade

How to use asharrm/juke-1.7b-dev-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull asharrm/juke-1.7b-dev-GGUF:Q4_K_M

Run and chat with the model

lemonade run user.juke-1.7b-dev-GGUF-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use asharrm/juke-1.7b-dev-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 asharrm/juke-1.7b-dev-GGUF: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 asharrm/juke-1.7b-dev-GGUF:Q4_K_M

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use asharrm/juke-1.7b-dev-GGUF with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf asharrm/juke-1.7b-dev-GGUF: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 "asharrm/juke-1.7b-dev-GGUF: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"

juke-1.7b v1.1: GGUF

GGUF builds of juke-1.7b v1.1, a 1.7B model that turns any prompt into a playlist of real songs. Type "bossa nova brunch" or "my cat's tax return" and get a title, a one-line idea, and a numbered tracklist. It can also think out loud like a DJ first.

Use these files with llama.cpp, Ollama, LM Studio, or anything else that reads GGUF. The chat template is included.

Which file?

File	Size	Name matches code	Repeats	Speed*	Notes
`juke-1.7b-v1.1-bf16.gguf`	3.45 GB	97.3%	1.7%	77 tok/s	Full precision reference
`juke-1.7b-v1.1-q8_0.gguf`	1.84 GB	98.0%	1.8%	151 tok/s	Same quality as bf16
`juke-1.7b-v1.1-q6_k.gguf`	1.42 GB	98.5%	1.6%	182 tok/s
`juke-1.7b-v1.1-q5_k_m.gguf`	1.26 GB	98.0%	2.2%	192 tok/s
`juke-1.7b-v1.1-q4_k_m.gguf`	1.11 GB	97.1%	1.3%	202 tok/s	Recommended

All quants write 100% valid song codes. Q3_K_M is not provided: in testing, the song name it wrote matched the song code only 74% of the time, so it names the wrong songs.

*Generation speed on a PS5 running Linux (AMD gfx1013 GPU, llama.cpp b11669, Vulkan, all layers on GPU).

How it was tested: juke's own probe (eval_overfit.py). 110 greedy generations with thinking on and off, about 1,000 tracks per file: held-out prompts, paraphrases, and 40 varied prompts. Name matches code = the artist and title the model wrote belong to the song code it wrote. Repeats = share of tracks that repeat inside one playlist.

Quick start

# llama.cpp
llama-server -hf asharrm/juke-1.7b-dev-GGUF:Q4_K_M -ngl 99 -c 2048
# then chat at http://127.0.0.1:8080

# Ollama
ollama run hf.co/asharrm/juke-1.7b-dev-GGUF:Q4_K_M

To turn off the DJ thinking, start the assistant turn with an empty think block (<think>\n\n</think>\n\n), or use /no_think in clients that support it.

Each track line looks like 1. <a_67><b_10><c_0> Brian Eno – An Ending — reason. The <a_><b_><c_> code identifies one of 229,042 songs in juke's catalog. See the main model card for the format, the training details and the v1.1 repeat fix.

License

CC-BY-SA-4.0, the same as the original model. Base model: Qwen/Qwen3-1.7B.

Downloads last month: 327

GGUF

Model size

2B params

Architecture

qwen3

Hardware compatibility

4-bit

5-bit

6-bit

8-bit

16-bit

Model tree for asharrm/juke-1.7b-dev-GGUF

Base model

Qwen/Qwen3-1.7B-Base

Finetuned

Qwen/Qwen3-1.7B

Quantized

asharrm/juke-1.7b-dev

Quantized

(1)

this model