ML Notes

Hugging Face and Sentence Transformers on M1 Macs

Easy guide to Hugging Face and Sentence-Transformers on M1 with GPU-acceleration

James Briggs

2 min readJun 3, 2022

Prerequisites

MacOS 12.4+
Anaconda distribution for M1 (find at bottom of page here)

Python Environment

We first setup a new Python environment that uses osx-arm64 and install the latest version(s) of PyTorch.

Open terminal
Create new environment with conda create -n ml python=3.9 anaconda
[Alternative]: If using the non-M1 distribution of Anaconda use CONDA_SUBDIR=osx-arm64 conda create -n ml python=3.9 -c conda-forge
Switch to new environment with conda activate ml
Set the sub-directory variable with conda env config vars set CONDA_SUBDIR=osx-arm64
Reactivate the environment with conda activate followed by conda activate ml

PyTorch Installation

GPU-acceleration on the new M1 chips is a new feature for PyTorch. For this, we need PyTorch v1.12 or later, as of 3rd June 2022 this is only available via the PyTorch nightly release.

pip install -U --pre torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/nightly/cpu

You should be able to see something like **Downloading torch-1.1x.x. — arm64.whl**. That final **arm64.whl** part is important and tells us we are downloading the correct version.

*Here’s a video on the new GPU-enabled PyTorch for M1*

Hugging Face Installation

The Transformers library uses another Hugging Face package called Tokenizers. Tokenizers requires Rust, because we’re using an ARM64 environment we will likely see ERROR: Failed building wheel for tokenizers during the install for these packages.

To avoid this, we install Rust.

curl — proto ‘=https’ — tlsv1.2 -sSf https://sh.rustup.rs | sh

Now we can pip install transformers.

Sentence Transformers Installation

Our final step is installing the Sentence Transformers library, again there are some additional steps we must take to get this working on M1.

Sentence transformers has a sentencepiece depency, if we try to install this package we will see ERROR: Failed building wheel for sentencepiece. To fix this we need:

brew install cmake
brew install pkg-config

*If you don’t have Homebrew installed click here*

Now we’re ready to pip install sentence-transformers.

And that’s it, you’re good to go!

If you’d like to keep up to date with what I’m doing I post weekly on YouTube, and you get in touch directly via Discord. I hope to see you around!