MacOS 12.2.1 Apple Silicon (M1 chip) with GPU acceleration

1. This was tested on macOS Monterey (12.2.1)

2. Install Miniforge for Apple Silicon. (See project page for more info.)

3. Create a new environment for pixplot:

conda create --name pixplot python=3.9

conda activate pixplot

4. Install the required packages for Apple's support of Tensorflow on the M1 chip:

conda install -c apple tensorflow-deps

pip install tensorflow-macos tensorflow-metal

5. Clone pixplot:

git clone https://github.com/YaleDHLab/pix-plot.git

6. Adjust the install_requires section of setup.py to read as follows:

  install_requires=[
    'cmake',
    'Cython',
    'glob2',
    'h5py',
    'iiif-downloader',
    'numpy',
    'pointgrid>=0.0.2',
    'python-dateutil>=2.8.0',
    'scikit-learn',
    'scipy',
    'six',
    'tqdm',
    'umap-learn',
    'yale-dhlab-rasterfairy>=1.0.3',
    'yale-dhlab-keras-preprocessing>=1.1.1',
    'matplotlib'
  ],

7. Fix Pillow:

pip install --force-reinstall https://files.pythonhosted.org/packages/34/4e/a491038f9bb13ea7f69874789d2cb69577eefb16724d678602f42222128f/Pillow-9.0.1-1-cp39-cp39-macosx_11_0_arm64.whl

8. Install numba via conda, not pip:

conda install -c numba numba

9. Run python setup.py install

10. Optional steps past this point:

pip install image_datasets (a helper utility to easily grab pixplot-ready datasets)

11. Get a sample dataset:

python

import image_datasets

image_datasets.oslomini.download()

(now exit python via quit() or CTRL+D)

12. Test the installation:

time pixplot --images="datasets/oslomini/images/*.jpg" --metadata="datasets/oslomini/metadata/metadata.csv" --shuffle