Can I use the pytorch.cuda documentation and code on an Apple Silicon Machine?

[ayushdeolasee]

Can I use the pytorch.cuda documentation and code on an Apple Silicon Machine?

I have an Apple silicon machine, and I installed the supported PyTorch packages, I have verified that the GPU is available and I am able to transfer some of the tensors to the GPU, however in the PyTorch documentation and in the course, the torch.cuda was used which is for Nvidia GPUs but can I transfer that same code to an Apple Silicon Device?

An example is in the exercises chapter in the book (First Chapter, PyTorch Fundamentals), there is a task to find the manual_seed equivalent to the GPU, however that was for the CUDA GPUs, How can I transfer that to Apple Silicon?

[mrdbourke]

Hi @pypdeveloper ,

torch.cuda is specifically targeted for NVIDIA devices with CUDA enabled.

For Apple Silicon devices, you'll want to use torch.mps (MPS = metal performance shaders).

See the documentation here: https://pytorch.org/docs/stable/notes/mps.html

Also for manual seed on MPS, you'll want to use torch.mps.manual_seed(42) (you can use any number, I just used 42 as an example), see the docs: https://pytorch.org/docs/stable/generated/torch.mps.manual_seed.html#torch.mps.manual_seed

To setup your device for Apple Silicon, you can use the example code:

import torch
import numpy as np
import pandas as pd
import sklearn
import matplotlib.pyplot as plt

print(f"PyTorch version: {torch.__version__}")

# Check PyTorch has access to MPS (Metal Performance Shader, Apple's GPU architecture)
print(f"Is MPS (Metal Performance Shader) built? {torch.backends.mps.is_built()}")
print(f"Is MPS available? {torch.backends.mps.is_available()}")

# Set the device      
device = "mps" if torch.backends.mps.is_available() else "cpu"
print(f"Using device: {device}")

I took the code from my GitHub on setting up PyTorch for Mac: https://github.com/mrdbourke/pytorch-apple-silicon