huggingface/transformers · error · TypeError

Expected config to be an ExecutorchConfig or dict, got {type

Error message

Expected config to be an ExecutorchConfig or dict, got {type(config)}

What it means

ExecutorchExporter.export accepts config only as an ExecutorchConfig instance or a plain dict of its fields (converted via ExecutorchConfig(**config)); the check is strict (type(config) is not ExecutorchConfig), so even subclasses of ExecutorchConfig are rejected. Anything else raises this TypeError before backend preparation begins.

Source

Thrown at src/transformers/exporters/exporter_executorch.py:136

    >>> et_program = exporter.export(model, inputs, config=ExecutorchConfig(backend="xnnpack"))
    >>> et_program.write_to_file("model.pte")
    ```
    """

    required_packages = ["torch", "executorch"]
    tested_versions = {"torch": "2.12.0", "executorch": "1.3.1"}

    def export(
        self,
        model: PreTrainedModel,
        sample_inputs: MutableMapping[str, Any],
        config: ExecutorchConfig | dict[str, Any],
    ) -> ExecutorchProgramManager:
        """Export a model to ExecuTorch, applying backend preparation and torch op patches."""
        if isinstance(config, dict):
            config = ExecutorchConfig(**config)
        elif type(config) is not ExecutorchConfig:
            raise TypeError(f"Expected config to be an ExecutorchConfig or dict, got {type(config)}")

        prepare_for_backend = _BACKEND_PREPARE.get(config.backend)
        if prepare_for_backend is None:
            raise ValueError(f"Unsupported backend {config.backend} for ExecuTorch export")

        model, sample_inputs, partitioner = prepare_for_backend(model, sample_inputs)

        with apply_patches("executorch"), apply_patches(f"executorch.{config.backend}"):
            exported_program: ExportedProgram = super().export(model, sample_inputs, config=config)
            apply_fx_program_fixes("executorch", exported_program)
            apply_fx_node_fixes("executorch", exported_program.graph_module)
            edge_program_manager: EdgeProgramManager = to_edge_transform_and_lower(
                exported_program, partitioner=partitioner, compile_config=_get_edge_compile_config()
            )
            executorch_programs_manager: ExecutorchProgramManager = edge_program_manager.to_executorch(
                config=_get_backend_config(config)
            )

View on GitHub (pinned to a597f97485)

Solutions

1. Pass an ExecutorchConfig: ExecutorchExporter().export(model, inputs, config=ExecutorchConfig(backend="xnnpack")).
2. For extra/custom fields, pass a dict: config={"backend": "xnnpack", ...my_extra_fields} — dicts are splatted into the constructor.
3. Verify with type(config) is ExecutorchConfig before the call if dispatching dynamically.

Example fix

# before
ExecutorchExporter().export(model, inputs, config=DynamoConfig())  # TypeError

# after
from transformers.exporters.exporter_executorch import ExecutorchConfig
ExecutorchExporter().export(model, inputs, config=ExecutorchConfig(backend="xnnpack"))

from transformers.exporters.exporter_executorch import ExecutorchConfig

if type(config) is not ExecutorchConfig:
    config = ExecutorchConfig(**config)  # dicts OK; subclasses/other types are rejected by the exporter
ExecutorchExporter().export(model, inputs, config)

def is_executorch_config_like(cfg) -> bool:
    from transformers.exporters.exporter_executorch import ExecutorchConfig
    return type(cfg) is ExecutorchConfig or isinstance(cfg, dict)

When it happens

Trigger: Passing a DynamoConfig or OnnxConfig to ExecutorchExporter.export; passing a subclass of ExecutorchConfig; passing a string/None/path in the config slot.

Common situations: Reusing one config object across backends in a multi-format export script; extending ExecutorchConfig with extra fields via subclassing (blocked by the exact-type check — use a dict instead).

Related errors

• Expected config to be a DynamoConfig or dict, got {type(conf
• Unsupported backend {config.backend} for ExecuTorch export
• Expected config to be an OnnxConfig or dict, got {type(confi
• out_indices must be a list, got {type(self._out_indices)}
• You can only update int, float, bool or string values in the

AI-assisted analysis of huggingface/transformers@a597f97485 (2026-08-14). Data as JSON: /api/errors/6e7ea4c8f2673c39. Report an issue: GitHub.