MiniForge & MiniConda on M1

August 23, 2022

Working with MacBook M1 which is shipped with ARM64 instead of Intel's x86–64 processors was superb fun.

Hold my beer, it is no fun when it comes to running Tensorflow on M1.

Why MiniForge & MiniConda

Here's the answer nerohmot from Stackoverflow,

miniforge is the community (conda-forge) driven minimalistic conda installer. Subsequent package installations come thus from conda-forge channel.

miniconda is the Anaconda (company) driven minimalistic conda installer. Subsequent package installations come from the anaconda channels (default or otherwise).

For several projects, we might tend to use different versions of python. One of the shortcomings of MiniForge is no back-port compatibility for older versions of Python. So we have to settle with python 3.8 where MiniConda helps us to run older versions on Mac M1 with the help of Rosetta.

If you are not using Homebrew, GET IT NOW bruh. It makes your life easier, a free and open-source software package management system that makes installation of any software on your macOS.

Installing Homebrew:

/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"

Installing MiniConda & MiniForge

For installing both MiniConda & MiniForge:

brew install miniforge

We cannot use Homebrew for both, as it will not allow since 'miniconda' conflicts with 'miniforge'. For 'miniconda' installation, we can go with the classic bash installation after getting the installer link from MiniConda. The following commands will make miniforge active on your machine.

wget https://repo.anaconda.com/miniconda/Miniconda3-latest-MacOSX-arm64.sh
chmod +x Miniforge3-MacOSX-arm64.sh
sh Miniforge3-MacOSX-arm64.sh
source ~/miniforge3/bin/activate

Installing Tensorflow

conda install -c apple tensorflow-deps
python -m pip install tensorflow-macos

Now we have TensorFlow successfully installed and we have miniconda for using older versions as well. For switching between miniconda and miniforge, we initialize conda with what we want. For conda initialization, I have tweaked the .zshrc.

You can also find the gist here.

Thanks for reading 👋 Happy Coding 👨‍💻

← back to the blog