python-accelerate fails to build with Python 3.15.0a8. 3 tests fail with a similar issue: ____________ CPUOptimizerTester.test_accelerated_optimizer_pickling ____________ self = <test_optimizer.CPUOptimizerTester testMethod=test_accelerated_optimizer_pickling> def test_accelerated_optimizer_pickling(self): model = torch.nn.Linear(10, 10) > optimizer = torch.optim.SGD(model.parameters(), 0.1) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ tests/test_optimizer.py:28: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ /usr/lib64/python3.15/site-packages/torch/optim/sgd.py:65: in __init__ super().__init__(params, defaults) /usr/lib64/python3.15/site-packages/torch/optim/optimizer.py:408: in __init__ self.add_param_group(cast(dict, param_group)) /usr/lib64/python3.15/site-packages/torch/_compile.py:47: in inner import torch._dynamo /usr/lib64/python3.15/site-packages/torch/_dynamo/__init__.py:76: in <module> from .polyfills import loader as _ # usort: skip # noqa: F401 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ /usr/lib64/python3.15/site-packages/torch/_dynamo/polyfills/loader.py:34: in <module> POLYFILLED_MODULES: tuple["ModuleType", ...] = tuple( ^^^^^ /usr/lib64/python3.15/site-packages/torch/_dynamo/polyfills/loader.py:35: in <genexpr> importlib.import_module(f".{submodule}", package=polyfills.__name__) /usr/lib64/python3.15/importlib/__init__.py:88: in import_module return _bootstrap._gcd_import(name[level:], package, level) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ /usr/lib64/python3.15/site-packages/torch/_dynamo/polyfills/struct.py:20: in <module> @substitute_in_graph(struct.pack, can_constant_fold_through=True) # type: ignore[arg-type] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ traceable_fn = <function pack at 0x7f50247c78a0> def wrapper(traceable_fn: Callable[_P, _R]) -> Callable[_P, _R]: if not is_function(traceable_fn): raise TypeError( f"@substitute_in_graph(...) expects a function but got {type(traceable_fn)!r}" ) if not skip_signature_check: try: original_sig = inspect.signature(original_fn) except ValueError: pass else: traceable_sig = inspect.signature(traceable_fn) def sig_ident( sig: inspect.Signature, ) -> tuple[tuple[str, ...], set[str], dict[str, Any]]: # Ignore annotations for parameters and return type return ( tuple( p.name for p in sig.parameters.values() if ( p.kind not in { p.KEYWORD_ONLY, # the name of *args and **kwargs is not important p.VAR_POSITIONAL, p.VAR_KEYWORD, } ) ), { p.name for p in sig.parameters.values() if p.kind == p.KEYWORD_ONLY }, { p.name: p.default for p in sig.parameters.values() # the name of *args and **kwargs is not important if p.kind not in {p.VAR_POSITIONAL, p.VAR_KEYWORD} }, ) wildcard_sig = inspect.signature(lambda *args, **kwargs: None) if ( sig_ident(original_sig) != sig_ident(traceable_sig) and sig_ident(original_sig) != sig_ident(wildcard_sig) and sig_ident(traceable_sig) != sig_ident(wildcard_sig) ): > raise TypeError( f"Signature mismatch between {original_fn} and {traceable_fn}: " f"{original_sig} != {traceable_sig}" ) E TypeError: Signature mismatch between <built-in function pack> and <function pack at 0x7f50247c78a0>: (format, /, *values) != (fmt: 'bytes | str', /, *v: 'Any') -> 'bytes' https://docs.python.org/3.15/whatsnew/3.15.html For the build logs, see: https://copr-be.cloud.fedoraproject.org/results/@python/python3.15/fedora-rawhide-x86_64/10341837-python-accelerate/ For all our attempts to build python-accelerate with Python 3.15, see: https://copr.fedorainfracloud.org/coprs/g/python/python3.15/package/python-accelerate/ Testing and mass rebuild of packages is happening in copr. You can follow these instructions to test locally in mock if your package builds with Python 3.15: https://copr.fedorainfracloud.org/coprs/g/python/python3.15/ Let us know here if you have any questions. Python 3.15 is planned to be included in Fedora 45. To make that update smoother, we're building Fedora packages with all pre-releases of Python 3.15. A build failure prevents us from testing all dependent packages (transitive [Build]Requires), so if this package is required a lot, it's important for us to get it fixed soon. We'd appreciate help from the people who know this package best, but if you don't want to work on this now, let us know so we can try to work around it on our side.
Hmm, thank you for alerting me to this. I wonder if this is actually an issue with PyTorch's dynamo not being compatible with Python 3.15? I'm adding Tom Rix, who is the torch maintainer, to see if this matches what he's seen previously.