MPS backend

Created On: May 11, 2026 | Last Updated On: Sep 09, 2026

mps device enables high-performance training on GPU for macOS devices with Metal programming framework. It introduces a new device to map Machine Learning computational graphs and primitives on highly efficient Metal Performance Shaders Graph framework and tuned kernels provided by Metal Performance Shaders framework respectively.

The new MPS backend extends the PyTorch ecosystem and provides existing scripts capabilities to setup and run operations on GPU.

To get started, simply move your Tensor and Module to the mps device:

# Check that MPS is available
if not torch.backends.mps.is_available():
    if not torch.backends.mps.is_built():
        print("MPS not available because the current PyTorch install was not "
              "built with MPS enabled.")
    else:
        print("MPS not available because the current macOS version is not 14.0+ "
              "and/or you do not have an MPS-enabled device on this machine.")

else:
    mps_device = torch.device("mps")

    # Create a Tensor directly on the mps device
    x = torch.ones(5, device=mps_device)
    # Or
    x = torch.ones(5, device="mps")

    # Any operation happens on the GPU
    y = x * 2

    # Move your model to mps just like any other device
    model = YourFavoriteNet()
    model.to(mps_device)

    # Now every call runs on the GPU
    pred = model(x)

Double-precision types

The MPS backend does not support torch.float64 (torch.double) or torch.complex128 (torch.cdouble) because Metal Shading Language has no double type:

>>> torch.ones(3, dtype=torch.float64, device="mps")
TypeError: Cannot convert a MPS Tensor to float64 dtype as the MPS framework
doesn't support float64. Please use float32 instead.

This is a fundamental limitation of Metal Shading Language rather than an unimplemented PyTorch operator. Consequently, PYTORCH_ENABLE_MPS_FALLBACK=1 does not apply: tensors with double-precision types cannot be allocated or copied to MPS in the first place. Tensor.to("mps") on a float64 tensor raises for the same reason, rather than downcasting implicitly.

A computation that genuinely requires double precision has to run on the CPU.