Loading…
Apple’s tensorflow-metal plug-in enables accelerating TensorFlow training on Macs by running TensorFlow on macOS GPUs through the Metal PluggableDevice. It requires macOS 12.0 or later, Python 3.9+ (and Apple silicon or AMD GPUs), and you install base TensorFlow followed by the tensorflow-metal plug-in in a Python virtual environment, per the setup steps. The page also covers verification steps, troubleshooting, and limitations such as no support for complex data types, no multi-GPU support, and cases where CPU may outperform GPU for small workloads.
We haven’t found TensorFlow Metal in the AI answers we’ve checked.