Running a local LLM on my AMD GPU was a lot easier than I expected

Unlocking more RAM for local models.