mlflow/mlflow · error · NotImplementedError

Tags are not available for Databricks managed datasets. Tags

Error message

Tags are not available for Databricks managed datasets. Tags are managed through Unity Catalog. Use Unity Catalog APIs to manage dataset tags.

What it means

The tags property is only implemented for MLflow-managed evaluation datasets. For Databricks-managed datasets, tags live in Unity Catalog, so MLflow raises NotImplementedError directing you to UC APIs. This prevents silently returning wrong or empty tag data.

Source

Thrown at mlflow/genai/datasets/evaluation_dataset.py:191

    @property
    def created_time(self) -> int | str | None:
        """The time the dataset was created."""
        if self._mlflow_dataset:
            return self._mlflow_dataset.created_time
        return self._databricks_dataset.create_time

    @property
    def create_time(self) -> int | str | None:
        """Alias for created_time (for backward compatibility with managed datasets)."""
        return self.created_time

    @property
    def tags(self) -> dict[str, Any] | None:
        """The tags for the dataset (MLflow only)."""
        if self._mlflow_dataset:
            return self._mlflow_dataset.tags
        raise NotImplementedError(
            "Tags are not available for Databricks managed datasets. "
            "Tags are managed through Unity Catalog. Use Unity Catalog APIs to manage dataset tags."
        )

    @property
    def experiment_ids(self) -> list[str]:
        """The experiment IDs associated with the dataset (MLflow only)."""
        if self._mlflow_dataset:
            return self._mlflow_dataset.experiment_ids
        return self._databricks_dataset.experiment_ids

    @property
    def schema(self) -> str | None:
        """The schema of the dataset."""
        if self._mlflow_dataset:
            return self._mlflow_dataset.schema
        return self._databricks_dataset.schema if self._databricks_dataset else None

View on GitHub (pinned to 6a27f2decc)

Solutions

Manage and read tags via Unity Catalog APIs (e.g. ALTER TABLE ... SET TAGS in SQL, or the Databricks SDK's UC table tagging APIs)
Branch on the backend: only access .tags when the dataset is MLflow-managed
Store custom metadata outside the dataset (e.g. in an MLflow experiment tag keyed by dataset name) if UC tagging isn't available to you

Example fix

// before
tags = eval_ds.tags  # NotImplementedError for Databricks datasets

// after
if eval_ds._mlflow_dataset:
    tags = eval_ds.tags
else:
    from databricks.sdk import WorkspaceClient
    tags = WorkspaceClient().tables.get(eval_ds.name).securable_kind  # use UC tag APIs / SQL: SHOW TBLPROPERTIES

Defensive patterns

Strategy: validation

Validate before calling

if eval_ds._databricks_dataset is not None:
    # tags unavailable; use Unity Catalog tag APIs
    from databricks.sdk import WorkspaceClient
    tags = None  # or fetch via UC: SHOW TBLPROPERTIES / SDK

Try / catch

try:
    tags = eval_ds.tags
except NotImplementedError:
    tags = fetch_uc_tags(eval_ds.name)  # Unity Catalog tag lookup

Prevention

Never read .tags on Databricks-backed datasets; manage tags in Unity Catalog
Write backend-aware helpers that branch on _mlflow_dataset/_databricks_dataset
Document to your team that dataset tags live in UC, not MLflow

When it happens

Trigger: Reading .tags on an EvaluationDataset backed by a Databricks (UC-managed) dataset, e.g. ds = mlflow.genai.datasets.get_dataset(name="catalog.schema.x"); ds.tags.

Common situations: Code that generically reads dataset tags for logging/audit across both MLflow and Databricks backends, or teams migrating from MLflow-native datasets to UC-managed ones.

Understand the failure class

Background: UnsupportedOperationException and "is not supported" errors: when a library deliberately refuses a call — this error's family across 30 libraries.

Related errors

AI-assisted analysis of mlflow/mlflow@6a27f2decc (2026-08-29). Data as JSON: /api/errors/9f6124841a50e945. Report an issue: GitHub.