deepset-ai/haystack · error · ValueError

Unknown component device type '{dict['type']}' in serialized

Error message

Unknown component device type '{dict['type']}' in serialized data

What it means

ComponentDevice.from_dict() deserializes a dict produced by to_dict(), which must have type set to 'single' or 'multiple'. Any other 'type' value means the serialized data is corrupt, hand-written incorrectly, or produced by an incompatible version.

Source

Thrown at haystack/utils/device.py:484

            return {"type": "multiple", "device_map": self._multiple_devices.to_dict()}
        # Unreachable
        raise AssertionError()

    @classmethod
    def from_dict(cls, dict: dict[str, Any]) -> "ComponentDevice":  # noqa:A002
        """
        Create a component device representation from a JSON-serialized dictionary.

        :param dict:
            The serialized representation.
        :returns:
            The deserialized component device.
        """
        if dict["type"] == "single":
            return cls.from_str(dict["device"])
        if dict["type"] == "multiple":
            return cls.from_multiple(DeviceMap.from_dict(dict["device_map"]))
        raise ValueError(f"Unknown component device type '{dict['type']}' in serialized data")


def _get_default_device() -> Device:
    """
    Return the default device for Haystack.

    Precedence:
        GPU > XPU > MPS > CPU. If PyTorch is not installed, only CPU is available.

    :returns:
        The default device.
    """
    try:
        torch_import.check()

        has_mps = (
            hasattr(torch.backends, "mps")
            and torch.backends.mps.is_available()

View on GitHub (pinned to e318778c9b)

Solutions

1. Set 'type' to 'single' with a 'device' string key, or 'multiple' with a 'device_map' dict.
2. Re-serialize the device/pipeline from a working ComponentDevice.to_dict() output.
3. Check for version mismatch between the Haystack that saved the data and the one loading it.

Example fix

// before
ComponentDevice.from_dict({"type": "gpu", "device": "cuda:0"})
// after
ComponentDevice.from_dict({"type": "single", "device": "cuda:0"})

def valid_device_dict(d):
    return d.get("type") in ("single", "multiple") and ("device" in d or "device_map" in d)

def is_serialized_device(d: dict) -> bool:
    return isinstance(d, dict) and d.get("type") in {"single", "multiple"}

try:
    dev = ComponentDevice.from_dict(d)
except ValueError:
    dev = ComponentDevice.from_str(d.get("device", "cpu"))

When it happens

Trigger: Calling ComponentDevice.from_dict({'type': 'gpu', 'device': 'cuda:0'}) or loading YAML/JSON where the device entry's 'type' key was edited or came from an older schema.

Common situations: Hand-editing serialized pipeline YAML; deserializing pipelines saved by a different Haystack version; typos like 'Single' or 'multi'.

Related errors

• Refusing to deserialize an OutputAdapter with unsafe=True wh
• Refusing to deserialize an OutputAdapter with custom filters
• Couldn't deserialize component '{name}' of class '{component
• Component '{name}' of type '{type(component).__name__}' has
• Component '{name}' of type '{type(component).__name__}' has

AI-assisted analysis of deepset-ai/haystack@e318778c9b (2026-08-30). Data as JSON: /api/errors/d2ed39b2a899072e. Report an issue: GitHub.