Lightning-AI/pytorch-lightning · error · ValueError

Device should be MPS, got {device} instead.

Error message

Device should be MPS, got {device} instead.

What it means

The DD:HH:MM:SS string for val_check_interval must have four integer components. After the format check (exactly 4 parts) passes, each part is coerced with int(); if any component is non-numeric (e.g. '0:a:0:0', '0.5:0:0:0', empty string part), the ValueError from int() is converted into this MisconfigurationException.

Source

Thrown at src/lightning/fabric/accelerators/mps.py:41

from lightning.fabric.accelerators.registry import _AcceleratorRegistry


class MPSAccelerator(Accelerator):
    """Accelerator for Metal Apple Silicon GPU devices.

    .. warning::  Use of this accelerator beyond import and instantiation is experimental.

    """

    @override
    def setup_device(self, device: torch.device) -> None:
        """
        Raises:
            ValueError:
                If the selected device is not MPS.
        """
        if device.type != "mps":
            raise ValueError(f"Device should be MPS, got {device} instead.")

    @override
    def teardown(self) -> None:
        pass

    @staticmethod
    @override
    def parse_devices(devices: Union[int, str, list[int]]) -> Optional[list[int]]:
        """Accelerator device parsing logic."""
        from lightning.fabric.utilities.device_parser import _parse_gpu_ids

        return _parse_gpu_ids(devices, include_mps=True)

    @staticmethod
    @override
    def get_parallel_devices(devices: Union[int, str, list[int]]) -> list[torch.device]:
        """Gets parallel devices for the Accelerator."""
        parsed_devices = MPSAccelerator.parse_devices(devices)

View on GitHub (pinned to 9fed5c27d2)

Solutions

1. Ensure all four components are plain integers: '0:0:0:1' not '0:0:0:1.5'
2. Construct from validated ints: f"{d}:{h}:{m}:{s}" with each an int
3. Use timedelta(seconds=float) or a dict {'seconds': 1.5} when fractional units are needed

Example fix

# before
trainer = Trainer(val_check_interval="0:0:0:1.5")

# after
from datetime import timedelta
trainer = Trainer(val_check_interval=timedelta(seconds=1.5))

def parse_interval(s: str):
    parts = s.split(":")
    if len(parts) != 4 or not all(p.lstrip("-").isdigit() for p in parts):
        raise ValueError("val_check_interval string must be 'DD:HH:MM:SS' with integers")
    return s

def is_valid_interval_str(v) -> bool:
    if not isinstance(v, str):
        return False
    parts = v.split(":")
    return len(parts) == 4 and all(p.isdigit() for p in parts)

When it happens

Trigger: Trainer(val_check_interval='0:0:thirty:0'); Trainer(val_check_interval='00:00:00:1.5'); a component with whitespace or a float like '0.5'; empty component '0::0:0'.

Common situations: Building the string from f-strings with unvalidated variables; float seconds intended (e.g. 1.5s) which the format does not support; copy-paste with a missing value leaving an empty slot.

Related errors

• You requested to find {num_devices} devices but only {len(av
• Device should be CUDA, got {device} instead.
• You requested to find {num_devices} devices but there are no
• You requested to find {num_devices} devices but this machine
• `name` must be a str, found {name}

AI-assisted analysis of Lightning-AI/pytorch-lightning@9fed5c27d2 (2026-08-28). Data as JSON: /api/errors/00be3ddc36b4cad9. Report an issue: GitHub.