MPS_ERROR_UNSUPPORTED_DTYPE pytorch type_error ai_generated true

RuntimeError: MPS backend does not support float64. Falling back to float32. Please cast your tensors to float32 explicitly.

ID: pytorch/mps-float64-cast-fallback

Also available as: JSON · Markdown · 中文
90%Fix Rate
88%Confidence
1Evidence
2023-05-01First Seen

Version Compatibility

VersionStatusIntroducedDeprecatedNotes
PyTorch 2.0.0 active — — —
PyTorch 2.1.0 active — — —
macOS 13 Ventura active — — —
macOS 14 Sonoma active — — —
Apple M1 active — — —
Apple M2 active — — —
Apple M3 active — — —

Root Cause

The Metal Performance Shaders (MPS) backend on Apple Silicon does not support float64 (double precision) tensors; any operation involving float64 tensors triggers a fallback to CPU or an error, causing performance degradation or crashes.

generic

中文

Apple Silicon 上的 Metal Performance Shaders (MPS) 后端不支持 float64（双精度）张量；任何涉及 float64 张量的操作会触发回退到 CPU 或导致错误，引起性能下降或崩溃。

Official Documentation

https://pytorch.org/docs/stable/notes/mps.html#un-supported-features

Workarounds

  1. 95% success Cast all tensors to float32 explicitly before moving to MPS device: tensor = tensor.float().to('mps'). This ensures compatibility with MPS backend. 
    Cast all tensors to float32 explicitly before moving to MPS device: tensor = tensor.float().to('mps'). This ensures compatibility with MPS backend.
  2. 90% success Set the default dtype to float32 at the start of the script: torch.set_default_dtype(torch.float32). This prevents accidental creation of float64 tensors from Python floats. 
    Set the default dtype to float32 at the start of the script: torch.set_default_dtype(torch.float32). This prevents accidental creation of float64 tensors from Python floats.
  3. 85% success Use a custom wrapper function that checks dtype and casts if needed: def to_mps(tensor): return tensor.float().to('mps') if tensor.is_floating_point() else tensor.to('mps') 
    Use a custom wrapper function that checks dtype and casts if needed: def to_mps(tensor): return tensor.float().to('mps') if tensor.is_floating_point() else tensor.to('mps')

中文步骤

  1. Cast all tensors to float32 explicitly before moving to MPS device: tensor = tensor.float().to('mps'). This ensures compatibility with MPS backend.
  2. Set the default dtype to float32 at the start of the script: torch.set_default_dtype(torch.float32). This prevents accidental creation of float64 tensors from Python floats.
  3. Use a custom wrapper function that checks dtype and casts if needed: def to_mps(tensor): return tensor.float().to('mps') if tensor.is_floating_point() else tensor.to('mps')

Dead Ends

Common approaches that don't work:

  1. Set environment variable PYTORCH_MPS_HIGH_WATERMARK_RATIO=0.0 to disable MPS fallback 95% fail

    This flag controls memory watermark, not dtype support. The float64 issue remains and will cause errors.

  2. Use torch.set_default_dtype(torch.float64) to force double precision 90% fail

    This makes the problem worse by creating more float64 tensors, increasing fallback frequency and memory usage.

  3. Install PyTorch nightly build for better MPS support 85% fail

    Nightly builds may have improvements but still do not support float64 on MPS as of current versions.