How to Setup TensorFlow on Apple M1 Pro and M1 Max (works for M1 too)

Setup a TensorFlow environment on Apple's M1 chips. We'll get TensorFlow to use the M1 GPU as well as install common data science and machine learning libraries using Conda.

person typing on Apple MacBook Pro 14-inch with M1 Pro chip

Let's get your M1 (normal, Pro or Max) setup for machine learning and data science.

Who is this blog post for?

You: have a new M1, M1 Pro, M1 Max machine and would like to get started doing machine learning and data science on it.

This post: teaches you how to install the most common machine learning and data science packages (TensorFlow, pandas, NumPy, Jupyter, matplotlib, scikit-learn) on your machine and make sure they run using sample code.

New or experienced?

If you're new to setting up environments and software packages, watch the video version alongside the longer text-based instructions below.

If you're experienced at setting up environments, the shorter text-based instructions should be enough.

Got an issue?

If you have issues, please post them on the GitHub Issues page so others can see.

Video version

You can find a step by step video version of this article on YouTube.

The code from the video is from my M1 machine learning speed test GitHub repo.


How to setup a TensorFlow environment on M1, M1 Pro, M1 Max using Miniforge (shorter version)

If you're experienced with making environments and using the command line, follow this version. If not, see the longer version below.

  1. Download and install Homebrew from https://brew.sh. Follow the steps it prompts you to go through after installation.
  2. Download Miniforge3 (Conda installer) for macOS arm64 chips (M1, M1 Pro, M1 Max).
  3. Install Miniforge3 into home directory.

Note: If you already have a version of Anaconda installed, it may cause conflicts when installing Miniforge (if you're using M1/Pro/Max, favour Miniforge because it's specifically designed for arm64 chips).

chmod +x ~/Downloads/Miniforge3-MacOSX-arm64.sh
sh ~/Downloads/Miniforge3-MacOSX-arm64.sh
source ~/miniforge3/bin/activate

4. Restart terminal.

5. Create a directory to setup TensorFlow environment.

mkdir tensorflow-test
cd tensorflow-test

6. Make and activate Conda environment with Python 3.8 (Python 3.8 is the most stable with M1/TensorFlow in my experience, though you could try with Python 3.x).

conda create --prefix ./env python=3.8 
conda activate ./env

7. Install TensorFlow dependencies from Apple Conda channel.

conda install -c apple tensorflow-deps

8. Install base TensorFlow (Apple's fork of TensorFlow is called tensorflow-macos).

python -m pip install tensorflow-macos

9. Install Apple's tensorflow-metal to leverage Apple Metal (Apple's GPU framework) for M1, M1 Pro, M1 Max GPU acceleration.

python -m pip install tensorflow-metal

10. (Optional) Install TensorFlow Datasets to run benchmarks included in this repo.

python -m pip install tensorflow-datasets

11. Install common data science packages.

conda install jupyter pandas numpy matplotlib scikit-learn

12. Start Jupyter Notebook.

jupyter notebook

13. Import dependencies and check TensorFlow version/GPU access.

import numpy as np
import pandas as pd
import sklearn
import tensorflow as tf
import matplotlib.pyplot as plt

# Check for TensorFlow GPU access
print(f"TensorFlow has access to the following devices:\n{tf.config.list_physical_devices()}")

# See TensorFlow version
print(f"TensorFlow version: {tf.__version__}")

If it all worked, you should see something like:

TensorFlow has access to the following devices:
[PhysicalDevice(name='/physical_device:CPU:0', device_type='CPU'),
PhysicalDevice(name='/physical_device:GPU:0', device_type='GPU')]
TensorFlow version: 2.5.0

How to setup a TensorFlow environment on M1, M1 Pro, M1 Max using Miniforge (longer version)

If you're new to creating environments, using a new M1, M1 Pro, M1 Max machine and would like to get started running TensorFlow and other data science libraries, follow the below steps.

Note: You're going to see the term "package manager" a lot below. Think of it like this: a package manager is a piece of software that helps you install other pieces (packages) of software.

Installing package managers (Homebrew and Miniforge)

  1. Download and install Homebrew from https://brew.sh. Homebrew is a package manager that sets up a lot of useful things on your machine, including Command Line Tools for Xcode, you'll need this to run things like git. The command to install Homebrew will look something like:
/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"

It will explain what it's doing and what you need to do as you go.

2. Download the most compatible version of Miniforge (minimal installer for Conda specific to conda-forge, Conda is another package manager and conda-forge is a Conda channel) from GitHub.

If you're using an M1 variant Mac, it's "Miniforge3-MacOSX-arm64" <- click for direct download.

Clicking the link above will download a shell file called Miniforge3-MacOSX-arm64.sh to your Downloads folder (unless otherwise specified).

Open Terminal.

We've now got a shell file capable of installing Miniforge, but to do so we'll have to modify it's permissions to make it executable.

To do so, we'll run the command chmod -x FILE_NAME which stands for "change mode of FILE_NAME to -executable".

We'll then execute (run) the program using sh.

chmod +x ~/Downloads/Miniforge3-MacOSX-arm64.sh
sh ~/Downloads/Miniforge3-MacOSX-arm64.sh

3. Activating Miniforge.

The above code should install Miniforge3 into your home directory (~/ stands for "Home" on Mac).

To check this, we can try to activate the (base) environment, we can do so using the source command.

source ~/miniforge3/bin/activate

If it worked, you should see something like the following in your terminal window.

(base) [email protected] ~ %

4. Restart terminal.

We've just installed some new software and for it to fully work, we'll need to restart terminal.

Creating a TensorFlow environment

Now we've got the package managers we need, it's time to install TensorFlow.

Let's setup a folder called tensorflow-test (you can call this anything you want) and install everything in there to make sure it's working.

Note: An environment is like a virtual room on your computer. For example, you use the kitchen in your house for cooking because it's got all the tools you need. It would be strange to have an oven in your bedroom. The same thing on your computer. If you're going to be working on specific software, you'll want it all in one place and not scattered everywhere else.
  1. Make a directory called tensorflow-test.

This is the directory we're going to be storing our environment. And inside the environment will be the software tools we need to run TensorFlow.

We can do so with the mkdir command which stands for "make directory".

mkdir tensorflow-test
Note: You could call the directory here anything you want, I'm just calling it tensorflow-test for this example. If you were working on a project to classify images of dogs you might make a folder called dog-vision.

2. Change into tensorflow-test.

For the rest of the commands we'll be running them inside the directory tensorflow-test so we need to change into it.

We can do this with the cd command which stands for "change directory".

cd tensorflow-test

3. Create a Conda environment.

Now we're inside the tensorflow-test directory, let's create a new Conda environment using the conda command (this command was installed when we installed Miniforge above).

We do so using conda create --prefix ./env which stands for "conda create an environment with the name file/path/to/this/folder/env". The . stands for "everything before".

For example, if I didn't use the ./env, my filepath looks like: /Users/daniel/tensorflow-test/env

conda create --prefix ./env

4. Activate the environment.

If conda created the environment correctly, you should be able to activate it using conda activate path/to/environment.

Short version:

conda activate ./env

Long version:

conda activate /Users/daniel/tensorflow-test/env
Note: It's important to activate your environment every time you'd like to work on projects that use the software you install into that environment. For example, you might have one environment for every different project you work on. And all of the different tools for that specific project are stored in its specific environment.

If activating your environment went correctly, your terminal window prompt should look something like:

(/Users/daniel/tensorflow-test/env) [email protected] tensorflow-test %

5. Installing the software we need (TensorFlow and other data science packages).

Now we've got a Conda environment setup, it's time to install the software we need.

Let's start by installing various TensorFlow dependencies (TensorFlow is a large piece of software and depends on many other pieces of software).

Rather than list these all out, Apple have setup a quick command so you can install almost everything TensorFlow needs in one line.

conda install -c apple tensorflow-deps

The above stands for "hey conda install all of the TensorFlow dependencies from the Apple Conda channel" (-c stands for channel).

If it worked, you should see a bunch of stuff being downloaded and installed for you.

6. Installing base TensorFlow

Now all of the TensorFlow dependencies have been installed, it's time install base TensorFlow.

Apple have created a fork (copy) of TensorFlow specifically for Apple Macs. It has all the features of TensorFlow with some extra functionality to make it work on Apple hardware.

This Apple fork of TensorFlow is called tensorflow-macos and is the version we'll be installing:

python -m pip install tensorflow-macos

Depending on your internet connection the above may take a few minutes since TensorFlow is quite a large piece of software.

7. Installing TensorFlow-Metal

Now we've got base TensorFlow installed, it's time to install tensorflow-metal.

Why?

Machine learning models often benefit from GPU acceleration. And the M1, M1 Pro and M1 Max chips have quite powerful GPUs.

TensorFlow allows for automatic GPU acceleration if the right software is installed.

And Metal is Apple's framework for GPU computing.

So Apple have created a plugin for TensorFlow (also referred to as a TensorFlow PluggableDevice) called tensorflow-metal to run TensorFlow on Mac GPUs.

We can install it using:

python -m pip install tensorflow-metal

If the above works, we should now be able to leverage our Mac's GPU cores to speed up model training with TensorFlow.

8. (Optional) Install TensorFlow Datasets.

Doing the above is enough to run TensorFlow on your machine. But if you'd like to run the benchmarks included in the M1 Machine Learning Test GitHub repo, you'll need TensorFlow Datasets.

TensorFlow Datasets provides a collection of common machine learning datasets to test out various machine learning code.

python -m pip install tensorflow-datasets

9. Install common data science packages.

If you'd like to run the benchmarks above or work on other various data science and machine learning projects, you're likely going to need Jupyter Notebooks, pandas for data manipulation, NumPy for numeric computing, matplotlib for plotting and Scikit-Learn for traditional machine learning algorithms and processing functions.

To install those in the current environment run:

conda install jupyter pandas numpy matplotlib scikit-learn

10. Start a Jupyter Notebook and test everything out.

To see if everything worked, try starting a Jupyter Notebook and importing the installed packages.

# Start a Jupyter notebook
jupyter notebook

Once the notebook is started, in the first cell:

import numpy as np
import pandas as pd
import sklearn
import tensorflow as tf
import matplotlib.pyplot as plt

# Check for TensorFlow GPU access
print(tf.config.list_physical_devices())

# See TensorFlow version
print(tf.__version__)

If it all worked, you should see something like:

TensorFlow has access to the following devices:
[PhysicalDevice(name='/physical_device:CPU:0', device_type='CPU'),
PhysicalDevice(name='/physical_device:GPU:0', device_type='GPU')]
TensorFlow version: 2.5.0

To see if it really worked, try running one of the notebooks in the M1 machine learning speed test GitHub repo end to end.


Results from machine learning benchmarks

I did a bunch of machine learning benchmarks on various Macs with Intel, M1, M1 Pro and M1 Max chips.

You can see them in the results directory of the M1 Machine Learning Speed Test GitHub.

There's also a blog post where I detail what happened in each of the experiments.

And again, if you have any issues with setup, please post them on the GitHub Issues page so others can see.