• Place deepak_resume_dataset.json in your working folder.
• Edit finetune_gemma_unsloth.py and set LOCAL_MODEL_PATH to your gemma files.
• Run python finetune_gemma_unsloth.py (watch VRAM; adjust 4bit / batch if necessary).
• After training, look in the gguf_model folder for .gguf.
• Create Modelfile that points to that .gguf file and run:
  • ollama create my-gemma -f Modelfile
  • ollama run my-gemma
• Test locally with ollama run or via Ollama GUI if installed.

Example Modelfile

Create a file named Modelfile (no extension):

FROM C:\full\path\to\gguf_model\gemma3_270m_q4_k_m.gguf
SYSTEM You are a resume assistant fine-tuned on a specific resume. Answer based on the resume content only.

Then ollama create gemma-resume -f Modelfile.