Offensivesec/ubuntu-support-llm

Ubuntu Support LLM

A small GPT2-based model (51M params) fine-tuned on Ubuntu Q&A dialogue. Refuses all non-Ubuntu questions.

Model Details

• —Architecture: GPT2 (512 embd, 8 layers, 8 heads)
• —Training data: sedthh/ubuntudialogueqa (~12k records)
• —Epochs: 5 | Loss: 2.076
• —Quantization: Q4KM (40 MB)

Run with Ollama

ollama create ubuntu-support -f Modelfile
ollama run ubuntu-support