pytorch compilation_error ai_generated true

torch.jit.frontend.UnsupportedNodeError: Set is not supported in TorchScript

ID: pytorch/torchscript-not-supported

Also available as: JSON · Markdown
80%Fix Rate
86%Confidence
3Evidence
2023-01-01First Seen

Version Compatibility

VersionStatusIntroducedDeprecatedNotes
2 active — — —

Root Cause

Python construct or PyTorch operation not supported in TorchScript. TorchScript supports a limited subset of Python.

generic

Workarounds

  1. 88% success Refactor unsupported Python features to TorchScript-compatible equivalents 
    Replace set() with List, dict comprehensions with loops, f-strings with string concatenation

    Sources: https://pytorch.org/tutorials/

  2. 85% success Use @torch.jit.ignore or @torch.jit.unused for non-exportable methods 
    @torch.jit.ignore
def debug_print(self, x): ...  # excluded from TorchScript, only callable in Python
  3. 82% success Switch to torch.compile() (PyTorch 2.0+) instead of TorchScript 
    model = torch.compile(model)  # supports full Python, better performance than TorchScript

Dead Ends

Common approaches that don't work:

  1. Use torch.jit.trace instead of torch.jit.script to avoid all errors 75% fail

    Trace only captures one execution path; control flow, dynamic shapes, and data-dependent branches are lost

  2. Wrap unsupported code in try/except to catch scripting errors 82% fail

    TorchScript compilation is ahead-of-time; try/except does not help with compilation failures