GGUF builds of Tekimax/granite-ml-coder — a compact Python / machine-learning coding assistant fine-tuned from ibm-granite/granite-3.1-1b-a400m-instruct. These run on CPU and Apple Silicon (Metal) via llama.cpp, Ollama, and LM Studio.

Files

File	Bits	Size	Notes
`granite-ml-coder-Q4_K_M.gguf`	4-bit	~378 MB	recommended — loads fast, near-full quality

Run it

Ollama (from the public registry)

ollama run tekimaxllc/granite-ml-coder "Write a sklearn pipeline for the iris dataset"

Ollama (from this GGUF directly)

ollama run hf.co/Tekimax/granite-ml-coder-GGUF

llama.cpp

bash

llama-cli -m granite-ml-coder-Q4_K_M.gguf \
  -p "Write a Keras autoencoder for network-traffic anomaly detection"

LM Studio

Search for Tekimax/granite-ml-coder-GGUF, download the Q4_K_M file, and chat.

Suggested system prompt

You are an expert Python machine-learning engineer. Write correct, runnable code,
explain each pipeline step, watch for overfitting and how gradient descent
converges, and recommend the best model or formula for the task.

Intended use & limitations

Good for drafting Python ML code (scikit-learn, pandas, NumPy, Keras) and explaining ML concepts, fully offline. It's a 1B model — treat output as a fast first draft and verify before use. See the full model card for training details and limitations.

License

Apache-2.0 (inherited from ibm-granite/granite-3.1-1b-a400m-instruct).