Good day, I’m currently doing R&D on image processing, and I stumbled upon an example that uses PyTorch. I’m currently in the process of installing PyTorch, and I’m wondering does PyTorch need an nVidia GPU? I’ve seen other image processing code that require CUDA, but CUDA requires an nVidia card to work. Right now, I’m on a MacBook pro and I have no access to a desktop with an nVidia card. I couldn’t exactly find a “System Requirements” page. So any help is very much appreciated. Thank you.
Some Mackbook Pro comes with nvidia GPUs. To check, see “About this mac” from the top left apple button. Even if you don’t have an nvidia GPU, you can still run pytorch in cpu only mode.
I understand, my MBP uses an Intel GPU. Does that mean I should select “None” in the CUDA version when picking the command to download PyTorch?
PyTorch can work in CPU mode as well, it will be (much) slower though.
In general there is very low support for OpenCL in any deep learning frameworks (Tensorflow, Theano, Chainer, Dynet, Mxnet, Caffe…), because Nvidia provided very efficient deep learning primitives. Researchers and library developers built on top of that. AMD/Intel are still playing catchup.
You can find some port of popular frameworks on github, especially by Hugh Perkins but you will probably spend a lot of time compiling from source, patching in upstream updates, etc. And then you will realize that you only have basic layers/loss functions available.
1 Like
@a.vill To answer your question, if your computer says you have an Intel GPU then yes you should select “None” for the CUDA version.
1 Like
@richard my system has Intel GPU and after selecting “None”, it still throws below error:
File “/home/****/lib/python3.7/site-packages/torch/cuda/init.py”, line 101, in _check_driver
http://www.nvidia.com/Download/index.aspx""")
AssertionError:
Found no NVIDIA driver on your system. Please check that you
have an NVIDIA GPU and installed a driver from
What command did you use to install pytorch?