myownipgit/transformers-on-macbook-m1-gpu

transformers-on-macbook-m1-gpu

Around 1.5-2x improvement in training time over m1 cpu.

Install Pytorch for Macbook M1 GPU

Step 1: Install Xcode

$ xcode-select --install

Step 2: Setup a new conda environment

$ conda create -n torch-gpu python=3.8
$ conda activate torch-gpu

Step 3: Install Pytorch

$ conda install pytorch torchvision torchaudio -c pytorch-nightly

# If not working with conda then try pip
$ pip3 install --pre torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/nightly/cpu

Step 4: Sanity Check

import torch
import math
# this ensures that the current MacOS version is at least 12.3+
print(torch.backends.mps.is_available())
# this ensures that the current current PyTorch installation was built with MPS activated.
print(torch.backends.mps.is_built())

Hugging Face transformers Installation

Step 1: Install Rust

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

Step 2: Install transformers

$ pip install transformers

Train QA model

$ sh run.sh

Benchmark

System	Training Time (200 steps)
Kaggle P100 Notebook	~32 sec
Colab T4 Notebook	~47 sec
M1 GPU (Macbook pro 13 2020 model)	~328 sec
M1 CPU (Macbook pro 13 2020 model)	~527 sec
Kaggle CPU Notebook	~1500 sec

Reference

1. https://pytorch.org
2. https://towardsdatascience.com/installing-pytorch-on-apple-m1-chip-with-gpu-acceleration-3351dc44d67c
3. https://jamescalam.medium.com/hugging-face-and-sentence-transformers-on-m1-macs-4b12e40c21ce