Instructions to use h0ney-badger/qwen2.5-coder-1.5b-python-distill with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use h0ney-badger/qwen2.5-coder-1.5b-python-distill with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="h0ney-badger/qwen2.5-coder-1.5b-python-distill",
	filename="qwen-coder-1.5b-py-Q4_K_M.gguf",
)

llm.create_chat_completion(
	messages = [
		{
			"role": "user",
			"content": "What is the capital of France?"
		}
	]
)

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use h0ney-badger/qwen2.5-coder-1.5b-python-distill 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 h0ney-badger/qwen2.5-coder-1.5b-python-distill:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf h0ney-badger/qwen2.5-coder-1.5b-python-distill:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf h0ney-badger/qwen2.5-coder-1.5b-python-distill:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf h0ney-badger/qwen2.5-coder-1.5b-python-distill: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 h0ney-badger/qwen2.5-coder-1.5b-python-distill:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf h0ney-badger/qwen2.5-coder-1.5b-python-distill: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 h0ney-badger/qwen2.5-coder-1.5b-python-distill:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf h0ney-badger/qwen2.5-coder-1.5b-python-distill:Q4_K_M

Use Docker

docker model run hf.co/h0ney-badger/qwen2.5-coder-1.5b-python-distill:Q4_K_M

LM Studio
Jan

vLLM

How to use h0ney-badger/qwen2.5-coder-1.5b-python-distill with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "h0ney-badger/qwen2.5-coder-1.5b-python-distill"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "h0ney-badger/qwen2.5-coder-1.5b-python-distill",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/h0ney-badger/qwen2.5-coder-1.5b-python-distill:Q4_K_M

Ollama
How to use h0ney-badger/qwen2.5-coder-1.5b-python-distill with Ollama:
```
ollama run hf.co/h0ney-badger/qwen2.5-coder-1.5b-python-distill:Q4_K_M
```

Unsloth Studio

How to use h0ney-badger/qwen2.5-coder-1.5b-python-distill 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 h0ney-badger/qwen2.5-coder-1.5b-python-distill 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 h0ney-badger/qwen2.5-coder-1.5b-python-distill to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for h0ney-badger/qwen2.5-coder-1.5b-python-distill to start chatting

How to use h0ney-badger/qwen2.5-coder-1.5b-python-distill with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf h0ney-badger/qwen2.5-coder-1.5b-python-distill: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": "h0ney-badger/qwen2.5-coder-1.5b-python-distill:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use h0ney-badger/qwen2.5-coder-1.5b-python-distill with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf h0ney-badger/qwen2.5-coder-1.5b-python-distill: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 h0ney-badger/qwen2.5-coder-1.5b-python-distill:Q4_K_M

Run Hermes

hermes

Atomic Chat new

OpenClaw new

How to use h0ney-badger/qwen2.5-coder-1.5b-python-distill with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf h0ney-badger/qwen2.5-coder-1.5b-python-distill: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 "h0ney-badger/qwen2.5-coder-1.5b-python-distill: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 h0ney-badger/qwen2.5-coder-1.5b-python-distill with Docker Model Runner:
```
docker model run hf.co/h0ney-badger/qwen2.5-coder-1.5b-python-distill:Q4_K_M
```

Lemonade

How to use h0ney-badger/qwen2.5-coder-1.5b-python-distill with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull h0ney-badger/qwen2.5-coder-1.5b-python-distill:Q4_K_M

Run and chat with the model

lemonade run user.qwen2.5-coder-1.5b-python-distill-Q4_K_M

List all available models

lemonade list

Qwen2.5-Coder-1.5B-Instruct — Python self-distill (GGUF)

A tiny, Python-focused QLoRA fine-tune of Qwen2.5-Coder-1.5B-Instruct, distilled from a Qwen2.5-Coder-14B-Instruct teacher on locally-generated, execution-verified Python data. Quantized to Q4_K_M — 941 MB — so it runs comfortably on CPU on old / low-power hardware (built to run on a 2014 ThinkPad W541, no GPU needed).

TL;DR — 81.5% Python pass@1 in under 1 GB. That's +7.4 points over the stock 1.5B, and it beats a stock 7B at Python while being ~5× smaller — plus it now writes complete, interactive programs, not just bare functions.

Results (execution-based pass@1, Python)

Same 54-sample held-out Python eval, same Q4_K_M quant. Base = the exact weights this was fine-tuned from.

Model	Python pass@1
Qwen2.5-Coder-1.5B-Instruct (base)	74.1% (40/54)
This model	81.5% (44/54)

For context, on the same eval the stock 7B scored 77.8% at Python — this 941 MB model edges it out for Python.

pass@1 = the model's code was executed against held-out tests and had to pass. Carries ±1–2 samples of sampling noise.

How it was made

Teacher (Qwen2.5-Coder-14B) generates Python tasks + solutions + tests → each is executed, only passing samples kept → QLoRA SFT of the 1.5B student (Unsloth, r=16, 3 epochs) → merged 16-bit → GGUF Q4_K_M. Same pipeline as the 7B sibling, Python-only.

Training data mixes two styles (~566 samples): execution-verified functions and complete, runnable programs from natural requests (calculators, CLIs, games, file tools — teacher-generated + hand-authored gold, each run-verified). The complete-program half is what makes it write whole interactive programs (using input(), menus, etc.) rather than bare functions.

⚠️ Run it right or it feels dumb

A 1.5B must be run with the chat template applied and low temperature, or it rambles. Use llama-server (applies the template automatically) or llama-cli -cnv with --temp 0.2. Do not use plain llama-cli -p "..." (raw completion, temp 0.8) — that's the usual reason a small local model seems broken.

Evaluation methodology

Execution-based pass@1 on a dedicated eval set disjoint from training (exact + fuzzy dedup). Deliberately not HumanEval/MBPP — the goal was an honest, contamination-controlled comparison against the base, not a leaderboard number.

Honest limitations

Python only. It was trained and evaluated on Python; don't expect other languages to benefit.
Modest, specialized gain. +3.7 points over an already-decent base, on a same-distribution eval — a neutral benchmark would likely show less.
Small model: fine for functions, scripts, and everyday Python help; not a reasoning-heavy or large-context coder.

Usage (CPU-friendly)

Q4_K_M GGUF, 941 MB. On CPU (e.g. an old laptop):

# llama.cpp on CPU — no GPU offload
llama-cli   -m qwen-coder-1.5b-py-Q4_K_M.gguf -p "Write a Python function to ..."
llama-server -m qwen-coder-1.5b-py-Q4_K_M.gguf -c 4096          # OpenAI-compatible API
# or load the .gguf in LM Studio

Provenance & license

License: Apache-2.0. Base (Qwen2.5-Coder-1.5B) and teacher (Qwen2.5-Coder-14B) are both Apache-2.0 — no restriction on training from model outputs — and the data is fully self-generated (no scraped corpus, no ToS-restricted API).
Base model: Qwen/Qwen2.5-Coder-1.5B-Instruct
Teacher: Qwen/Qwen2.5-Coder-14B-Instruct
Full pipeline: https://github.com/h0n3y-badger/code-distill

A companion to the 7B Python/C distill — see the repo for the reproducible pipeline.