How to run on gpu in apple M1

The (not so) new apple M1 chip has integrated GPU cores. I was trying running a simple PyTorch code on GPU on this machine, but torch.cuda.is_available(): returns False. I was wondering if this is because of PyTorch not supporting this GPU or is there any additional step that I should take before running?

1 Like

Hi,

The GPU part of the M1 chip is not based on nvidia CUDA so it won’t work with our CUDA backend, but The CPU part should run just fine.

1 Like

Thanks for the response! @albanD Is there a work in progress/plan to support M1 chip GPUs later?

OK, so for the folks ending up here, here is the latest update: GPU acceleration for Apple's M1 chip? · Issue #47702 · pytorch/pytorch · GitHub

2 Likes