huggingface/transformers · error · TypeError

Missing required field - '{f.name}'

Error message

Missing required field - '{f.name}'

What it means

TypeError raised by the strict dataclass init wrapper when a required field (no default, no default_factory) was not supplied. The wrapper iterates all declared fields and fails on the first missing mandatory one, naming it explicitly.

Source

Thrown at src/transformers/configuration_utils.py:129

        dataclass_fields = {f.name for f in fields(cls)}
        standard_kwargs = {k: v for k, v in kwargs.items() if k in dataclass_fields}

        # We need to call bare `__init__` without `__post_init__` but the `original_init` of
        # any dataclas contains a call to post-init at the end (without kwargs)
        if len(args) > 0:
            raise ValueError(
                f"{cls.__name__} accepts only keyword arguments, but found `{len(args)}` positional args."
            )

        for f in fields(cls):  # type: ignore
            if f.name in standard_kwargs:
                setattr(self, f.name, standard_kwargs[f.name])
            elif f.default is not MISSING:
                setattr(self, f.name, f.default)
            elif f.default_factory is not MISSING:
                setattr(self, f.name, f.default_factory())
            else:
                raise TypeError(f"Missing required field - '{f.name}'")

        # Pass any additional kwargs to `__post_init__` and let the object
        # decide whether to set the attr or use for different purposes (e.g. BC checks)
        additional_kwargs = {}
        for name, value in kwargs.items():
            if name not in dataclass_fields:
                additional_kwargs[name] = value

        self.__post_init__(**additional_kwargs)

    cls.__init__ = __init__
    return cls


@dataclass_transform(kw_only_default=True)
@strict(accept_kwargs=True)
@dataclass(repr=False)
class PreTrainedConfig(PushToHubMixin, RotaryEmbeddingConfigMixin, HeterogeneousConfigMixin):

View on GitHub (pinned to a597f97485)

Solutions

1. Supply the missing field named in the message: MyConfig(hidden_size=768, num_heads=12)
2. Check for typos in keyword names — a misspelled kwarg does not fill the field it was meant for
3. If upgrading transformers, diff the config class fields for newly required members

Example fix

# before
cfg = MyConfig(hidden_size=768)  # num_heads required
# after
cfg = MyConfig(hidden_size=768, num_heads=12)

from dataclasses import fields, MISSING
required = [f.name for f in fields(MyConfig) if f.default is MISSING and f.default_factory is MISSING]
missing = [r for r in required if r not in params]
assert not missing, f'supply required config fields: {missing}'
cfg = MyConfig(**params)

def has_all_required(cls, params: dict) -> bool:
    return all(
        f.name in params or f.default is not MISSING or f.default_factory is not MISSING
        for f in fields(cls)
    )

try:
    cfg = MyConfig(**params)
except TypeError as e:
    m = re.search(r"Missing required field - '(\w+)'", str(e))
    if m:
        raise ValueError(f'config needs value for {m.group(1)!r}') from e
    raise

When it happens

Trigger: Constructing a strict config while omitting a field declared without a default, e.g. MyConfig(hidden_size=768) when num_heads is also required. Also when a typo'd keyword silently lands in additional kwargs instead of the field.

Common situations: New config versions adding required fields, breaking older construction code; copy-pasting a partial config from docs; typos (num_head vs num_heads) that bypass field matching and leave the real field unset.

Related errors

• {cls.__name__} accepts only keyword arguments, but found `{l
• out_indices must be a list, got {type(self._out_indices)}
• out_indices must be valid indices for stage_names {self.stag
• out_indices must not contain any duplicates, got {self._out_
• out_indices must be in the same order as stage_names, expect

AI-assisted analysis of huggingface/transformers@a597f97485 (2026-08-14). Data as JSON: /api/errors/4909db87b8663ae6. Report an issue: GitHub.