ParentFull threadthangngoc89·Correct. You need CUDA, or ROCm, MPS (native to macOS) backends for running deep learning. I found it relatively easy to train some Pytorch model on beefy server with CUDA and running interference on my Macbook Air.View on HN