MLX

Fine-Tuning Gemma 4 on a Mac — A Colab-Free Training Setup with Apple Silicon + MLX

A practical walkthrough of fine-tuning Gemma 4 locally on Apple Silicon using MLX — no Google Colab required. Tested on M3 Max with 96GB unified memory, with memory sizing guidance, learning-rate tuning, overnight-run setup, adapter fusing for Ollama, and fixes for the errors you are most likely to hit.