How to convert Keras model to CoreML package on MacOS and use it.

0. Precondition

• Currently (2026. Nov), to install tensorflow on the MacOS, it needs Python 3.12. Higher version returns Errors.

• Pyenv : Version manager, it helps to choose Python version from among several versions.

> brew install pyenv

> echo 'eval "$(pyenv init -)" ~/.zshrc

> source ~/.zshrc

• Install Python 3.12 globally not locally

> pyenv install 3.12

> pyenv global 3.12

1. Install Tensorflow

> python3 -m pip install --upgrade pip

> python3 -m pip install tensorflow

2. Write Python code to get a model (fashion_mnist)

Python Code Link

• Write python codes (fashion_mnist_download.py) with gist link

• run

> python fashion_mnist_download.py

• "fashion_mnist.keras" file will be generated.

3. Install CoreMLTools

> pip install coremltools

4. Convert

convert.py

> python convert.py

• Output file: MyFashion.mlpackage

5. Import mlpackage file into the Xcode project

• Drag & Drop mlpackage file in the left project view.

• Xcode automatically generate MyFashion.swift

Keep updating...