Contributed by Hong-Ye Hu
.. warning::
This page is deprecated. Please visit the update tutorial <development_Mac.html>_ for the latest information.
The key issue addressed in this document is how to install both TensorFlow and Jax on a M2 chip MacOS without conflict.
Need graphical access to the machine.
Run xcode-select --install to install if on optimal internet.
Or Download from Apple Command Line Tools installation image then install if internet connection is weak.
Due to the limitation of MacOS and packages, the lastest version of python does not always function as desired, thus miniconda installation is advised to solve the issues. And use anaconda virtual environment is always a good habit.
curl -o ~/miniconda.sh https://repo.anaconda.com/miniconda/Miniconda3-latest-MacOSX-arm64.sh
bash ~/miniconda.sh -b -p $HOME/miniconda
source ~/miniconda/bin/activate
First, create a virtual environment, and make sure the python version is 3.8.5 by
conda create --name NewEnv python==3.8.5
conda activate NewEnv
Then, install the TensorFlow from .whl file (file can be downloaded from this URL). This will install TensorFlow version 2.4.1
pip install ~/Downloads/tensorflow-2.4.1-py3-none-any.whl
Next, one need to install Jax and Optax by
conda install jax==0.3.0
conda install optax==0.1.4
Now, hopefully, you should be able to use both Jax and TensorFlow in this environment. But sometimes, it may give you an error "ERROR: package Chardet not found.".
If that is the case, you can install it by conda install chardet.
Lastly, install tensorcircuit
pip install tensorcircuit
This is the solution that seems to work for M2-chip MacOS. Please let me know if there is a better solution!