⚠️ Disclaimer: Model Limitations & Retraining Plans

While this experiment aimed to explore the feasibility of small, local AI assistants, the current model struggles with generalization and often reinforces patterns from training data rather than adapting dynamically.

To address this, we will repeat the fine-tuning process, refining the dataset and training approach to improve response accuracy and adaptability.

The goal remains the same: a reliable, privacy-first AI assistant that runs locally on edge devices.

ethicalabs/Kurtis-Qwen2.5-0.5B-Instruct-DPO-GGUF

This model was converted to GGUF format from ethicalabs/Kurtis-Qwen2.5-0.5B-Instruct-DPO using llama.cpp via the ggml.ai's GGUF-my-repo space. Refer to the original model card for more details on the model.

Use with llama.cpp

Install llama.cpp through brew (works on Mac and Linux)

brew install llama.cpp

Invoke the llama.cpp server or the CLI.

CLI:

llama-cli --hf-repo ethicalabs/Kurtis-Qwen2.5-0.5B-Instruct-DPO-GGUF --hf-file kurtis-qwen2.5-0.5b-instruct-dpo-q4_k_m.gguf -p "The meaning to life and the universe is"

Server:

llama-server --hf-repo ethicalabs/Kurtis-Qwen2.5-0.5B-Instruct-DPO-GGUF --hf-file kurtis-qwen2.5-0.5b-instruct-dpo-q4_k_m.gguf -c 2048

Note: You can also use this checkpoint directly through the usage steps listed in the Llama.cpp repo as well.

Step 1: Clone llama.cpp from GitHub.

git clone https://github.com/ggerganov/llama.cpp

Step 2: Move into the llama.cpp folder and build it with LLAMA_CURL=1 flag along with other hardware-specific flags (for ex: LLAMA_CUDA=1 for Nvidia GPUs on Linux).

cd llama.cpp && LLAMA_CURL=1 make

Step 3: Run inference through the main binary.

./llama-cli --hf-repo ethicalabs/Kurtis-Qwen2.5-0.5B-Instruct-DPO-GGUF --hf-file kurtis-qwen2.5-0.5b-instruct-dpo-q4_k_m.gguf -p "The meaning to life and the universe is"

or

./llama-server --hf-repo ethicalabs/Kurtis-Qwen2.5-0.5B-Instruct-DPO-GGUF --hf-file kurtis-qwen2.5-0.5b-instruct-dpo-q4_k_m.gguf -c 2048