mlflow/mlflow · error · MlflowException

INVALID_PARAMETER_VALUE

INVALID_PARAMETER_VALUE

Error message

The specified Hugging Face dataset does not contain the specified targets column '{targets}'.

What it means

HuggingFaceDataset construction raises INVALID_PARAMETER_VALUE when a `targets` column name is given that does not exist in the datasets.Dataset's column_names. MLflow validates target presence at construction time so evaluation can group inputs/outputs later.

Source

Thrown at mlflow/data/huggingface_dataset.py:51

        source: HuggingFaceDatasetSource,
        targets: str | None = None,
        name: str | None = None,
        digest: str | None = None,
    ):
        """
        Args:
            ds: A Hugging Face dataset. Must be an instance of `datasets.Dataset`.
                Other types, such as :py:class:`datasets.DatasetDict`, are not supported.
            source: The source of the Hugging Face dataset.
            targets: The optional name of the Hugging Face dataset column containing targets
                (labels) for supervised learning.
            name: The name of the dataset. E.g. "wiki_train". If unspecified, a name is
                automatically generated.
            digest: The digest (hash, fingerprint) of the dataset. If unspecified, a digest
                is automatically computed.
        """
        if targets is not None and targets not in ds.column_names:
            raise MlflowException(
                f"The specified Hugging Face dataset does not contain the specified targets column"
                f" '{targets}'.",
                INVALID_PARAMETER_VALUE,
            )

        self._ds = ds
        self._targets = targets
        super().__init__(source=source, name=name, digest=digest)

    def _compute_digest(self) -> str:
        """
        Computes a digest for the dataset. Called if the user doesn't supply
        a digest when constructing the dataset.
        """
        df = next(
            self._ds.to_pandas(
                batch_size=_MAX_ROWS_FOR_DIGEST_COMPUTATION_AND_SCHEMA_INFERENCE, batched=True
            )

View on GitHub (pinned to 6a27f2decc)

Solutions

1. Set targets to one of ds.column_names (print them to verify).
2. Load the split that contains the targets column.
3. Pass targets=None if the dataset is inputs-only.

Example fix

// before
ds = load_dataset("imdb", split="test[0:100]")
ds_meta = mlflow.data.from_huggingface(ds, targets="target")
// after
ds_meta = mlflow.data.from_huggingface(ds, targets="label")  # ds.column_names includes 'label'

if targets is not None and targets not in ds.column_names:
    raise ValueError(
        f"targets column {targets!r} not in dataset columns {ds.column_names}"
    )

def has_column(ds, col: str) -> bool:
    return col is None or col in ds.column_names

from mlflow.exceptions import MlflowException
try:
    meta = mlflow.data.from_huggingface(ds, targets=targets)
except MlflowException as e:
    if "does not contain the specified targets column" in str(e):
        meta = mlflow.data.from_huggingface(ds, targets=None)
    else:
        raise

When it happens

Trigger: Calling mlflow.data.from_huggingface(ds, targets="label") where ds.column_names does not contain "label" (typo, wrong split, dataset without labels).

Common situations: Typo in the targets column name; loading a test split that lacks the label column present in train; datasets where the label column is named differently (e.g. "labels", "target", "answer").

Related errors

• INVALID_PARAMETER_VALUE
• INTERNAL_ERROR
• Found duplicated arguments in `HuggingFaceDatasetSource` and
• INVALID_PARAMETER_VALUE
• Failed to load base model '{effective_base_model}'. If the m

AI-assisted analysis of mlflow/mlflow@6a27f2decc (2026-08-29). Data as JSON: /api/errors/7d7c9dbe9aa5c5e4. Report an issue: GitHub.