Apple GPU MLX example

[chadbrewbaker]

I made a single python file example of a port to Apple GPU with the MLX library. Please give me some feedback. It's been a while since I programmed centroidal Voronoi codes. In practice I use these for time series vectors that have been normalized to (0,1).

Please give me some feedback.

[chadbrewbaker]

I would add a GitHub worker for macos-26-arm64 that can be ran manually - it should have an Apple GPU to test with. I'm not sure what GitHub workers have NVIDIA GPU available. Probably have to install UV or modify to pip.

uv run examples/mlxtest.py

Put this in .github/workflows:

name: Manual MLX Benchmark

# Triggered only manually via the GitHub Actions UI
on:
  workflow_dispatch:

jobs:
  benchmark:
    # 'macos-15' is the latest Apple Silicon (M1) runner in the docs.
    # It provides the arm64 architecture required for MLX.
    runs-on: macos-15

    steps:
      - name: Checkout code
        uses: actions/checkout@v4

      - name: Install uv
        uses: astral-sh/setup-uv@v5
        with:
          enable-cache: true

      - name: Run Benchmark
        # 'uv run' detects the /// script block in your file 
        # and automatically installs mlx, torch, and numpy.
        run: |
          uv run examples/mlxtest.py

[andy-yang-1]

@chadbrewbaker Thanks for your contribution! For now we don't have apple GPU so it is hard for us to test it. So merging this will take more time.