[breckenedge]

Macs have GPUs and their architecture means that GPU has access to the full system RAM. Cuda isn’t a requirement for running ML workloads on a GPU.