Lightning-AI/pytorch-lightning · error · ValueError

The start method '{self._start_method}' is not available on

Error message

The start method '{self._start_method}' is not available on this platform. Available methods are: {', '.join(mp.get_all_start_methods())}

What it means

The _MultiProcessingLauncher validates its start_method against multiprocessing.get_all_start_methods() at construction time. If the requested start method (e.g. 'forkserver' or 'spawn') is not available on the current platform (notably Windows only supports 'spawn'), it raises immediately.

Source

Thrown at src/lightning/pytorch/strategies/launchers/multiprocessing.py:79

          before calling ``Trainer.fit``.

    Args:
        strategy: A reference to the strategy that is used together with this launcher.
        start_method: The method how to start the processes.
            - 'spawn': The default start method. Requires all objects to be pickleable.
            - 'fork': Preferable for IPython/Jupyter environments where 'spawn' is not available. Not available on
              the Windows platform for example.
            - 'forkserver': Alternative implementation to 'fork'.

    """

    def __init__(
        self, strategy: "pl.strategies.ParallelStrategy", start_method: Literal["spawn", "fork", "forkserver"] = "spawn"
    ) -> None:
        self._strategy = strategy
        self._start_method = start_method
        if start_method not in mp.get_all_start_methods():
            raise ValueError(
                f"The start method '{self._start_method}' is not available on this platform. Available methods are:"
                f" {', '.join(mp.get_all_start_methods())}"
            )
        self.procs: list[mp.Process] = []
        self._already_fit = False

    @property
    @override
    def is_interactive_compatible(self) -> bool:
        # The start method 'spawn' is not supported in interactive environments
        # The start method 'fork' is the only one supported in Jupyter environments, with constraints around CUDA
        # initialization. For more context, see https://github.com/Lightning-AI/pytorch-lightning/issues/7550
        return self._start_method == "fork"

    @override
    def launch(self, function: Callable, *args: Any, trainer: Optional["pl.Trainer"] = None, **kwargs: Any) -> Any:
        """Launches processes that run the given function in parallel.

View on GitHub (pinned to 9fed5c27d2)

Solutions

1. Use start_method='spawn' (the default, available everywhere), or omit the argument
2. If you must pick a method, choose from multiprocessing.get_all_start_methods() on the target platform
3. On Windows, restructure code to be spawn-safe (guard with if __name__ == '__main__' and picklable top-level functions)

Example fix

# before
launcher = _MultiProcessingLauncher(strategy, start_method="fork")  # on Windows

# after
import multiprocessing as mp
launcher = _MultiProcessingLauncher(strategy, start_method=next(iter(sorted(mp.get_all_start_methods() & {"fork", "forkserver", "spawn"})), "spawn"))

import multiprocessing as mp
method = start_method if start_method in mp.get_all_start_methods() else "spawn"

When it happens

Trigger: Constructing a launcher (indirectly via a strategy that uses multiprocessing, e.g. DDPStrategy with launchers) with start_method='fork' or 'forkserver' on Windows or macOS where that method is unavailable; explicitly instantiating _MultiProcessingLauncher(strategy, start_method='fork') on Windows.

Common situations: Code developed on Linux (fork works) run on Windows; notebooks/scripts hardcoding fork for performance; older tutorials passing start_method explicitly.

Related errors

• The start method '{self._start_method}' is not available on
• Cannot re-initialize CUDA in forked subprocess. To use CUDA
• Lightning can't create new processes if CUDA is already init
• Launching multiple processes with the 'spawn' start method r
• Device should be CPU, got {device} instead.

AI-assisted analysis of Lightning-AI/pytorch-lightning@9fed5c27d2 (2026-08-28). Data as JSON: /api/errors/ad7dabebab75e6cf. Report an issue: GitHub.