I recently got a new MacBook Pro with the M1 chip. Since Apple switched from Intel to ARM-based processors, I had to go through some hoops to get my development environment set up. Here are a few notes on the setup processes, but note that much of this may go out-of-date soon.
You will need this to run many non-native apps. Run softwareupdate --install-rosetta in your terminal.
Spotify
Does not run natively. Install Rosetta first, then download normally from the Spotify website.
Productivity tools
Application
Notes
NordVPN
Does not run natively. Install Rosetta first, then download normally from the App Store.
Spark (email client)
Does not run natively. Install Rosetta first, then download normally from the App Store.
RescueTime
Does not run natively. Install Rosetta first, then download normally from the App Store.
Bear notes
Seems to run natively; download normally from the App Store.
Fantastical
Seems to run natively; download normally from the App Store.
Developer essentials
Utility
Notes
xcode command line utils
Run xcode-select --install in your terminal. Alternatively the terminal seems to prompt the user to download the xcode command line utils after you try to execute a command such as git.
zsh config
Looks like the default shell is zsh in MacOS Big Sur. I installed ohmyzsh normally using curl and it worked fine.
Release for Apple Silicon here. They say it's partial but it has worked for everything I needed. However, you must check out to master by additionally running the command git -C $(brew —repo homebrew/core) checkout master after you install brew.
node
Install brew as above, then run brew install node in your terminal.
yarn
Install brew as above, then run brew install yarn in your terminal.
python
Do not install with brew! Use miniforge, as this will make it infinitely easier to install numpy and other scientific computing libraries. Download the ARM64 version from here, then run bash Miniforge3-MacOSX-arm64.sh. Then you can create environments as you would using conda (i.e. conda create -n myenv python=3.8) and conda installnumpy/scipy whatever library of your choosing.
MacTeX
Believe it or not, there is a stable release for Apple Silicon here.
Engineering tools
Tool
Notes
Docker Desktop
Install Rosetta first, then download the (stable) M1 version from here. You need Rosetta because some binaries are still Darwin/AMD. Note that not all images are available for ARM64.
Postman
Does not run natively. Install Rosetta first, then download normally from the website.
Postgres and psycopg2
This one is really tricky! After lots of searching, I found working instructions in this comment. Install postgres via brew with brew install postgresql, then run brew link openssl and make sure you execute the exports in the output from this command. Then you should be able to pip install psycopg2-binary fine. If you're using psycopg2 in docker, write your images based off python:3.9 instead of slim images (see this comment).
gatsby
Make sure you have node and yarn installed. Gatsby uses libvips which you will have to compile and install globally yourself. Follow steps in this issue to do this. Basically, you use Mac ports to install specific dependencies.
tmux
Install brew as above, then run brew install tmux.
c/c++
Covered by xcode command line tools. To verify, run clang++ --version in your terminal.
grpcio
Install conda (miniforge) as above. When you create a conda environment, you can run conda install LIB_NAME pretty smoothly.
To leverage the M1 chip (GPU and CPU), you will need to download Apple's tensorflow_macos. I successfully did this in a conda environment by following the instructions in this issue. Basically, if you have conda (miniforge), you can write an environment.yml file based off this file and create a conda environment with this file. pip install the wheels as described in the instructions and you should be good to go.
To set the device to gpu, run from tensorflow.python.compiler.mlcompute import mlcompute; mlcompute.set_mlc_device(device_name='gpu'). It will give a warning that gpu does not work in eager mode, but I was able to train a neural network blazingly fast (compared to my Intel MacBook), and Activity Monitor showed GPU utilization.
pytorch
The pytorch team has not released a stable native version for Apple Silicon yet. The most hassle-free solution I found is to open the terminal with Rosetta and download cpu versions of torch and torchvision. However, if you really want to use your Apple Silicon and are happy with torch alone, the miniforge version of conda allows you to conda install pytorch or build from source using a Mac ARM64 version listed here. Some people have successfully built torch and torchvision from source following the steps in this issue, but I am too lazy to check if it works for me.
Microsoft Excel
Stable release for Apple Silicon can be downloaded through the Microsoft website or whatever university or organization allows you to download Microsoft Office.
Feel free to make a PR with edits to this README if you're interested in adding or updating instructions.
Note that the project description data, including the texts, logos, images, and/or trademarks,
for each open source project belongs to its rightful owner.
If you wish to add or remove any projects, please contact us at [email protected].
