mlflow/mlflow · error · MlflowException

INVALID_PARAMETER_VALUE

INVALID_PARAMETER_VALUE

Error message

Failed to find answer similarity metric for version {metric_version}. Please check the version

What it means

answer_similarity() imports mlflow.metrics.genai.prompts.<metric_version>.AnswerSimilarityMetric by class name derived from metric_version. If that module does not exist (ModuleNotFoundError), MLflow raises INVALID_PARAMETER_VALUE telling you to check the version.

Source

Thrown at mlflow/metrics/genai/metric_definitions.py:69

            the default parameters defined in the metric implementation.
        extra_headers: (Optional) Dictionary of extra headers to be passed to the judge model.
        proxy_url: (Optional) Proxy URL to be used for the judge model. This is useful when the
            judge model is served via a proxy endpoint, not directly via LLM provider services.
            If not specified, the default URL for the LLM provider will be used
            (e.g., https://api.openai.com/v1/chat/completions for OpenAI chat models).
        max_workers: (Optional) The maximum number of workers to use for judge scoring.
            Defaults to 10 workers.

    Returns:
        A metric object
    """
    if metric_version is None:
        metric_version = _get_latest_metric_version()
    class_name = f"mlflow.metrics.genai.prompts.{metric_version}.AnswerSimilarityMetric"
    try:
        answer_similarity_class_module = _get_class_from_string(class_name)
    except ModuleNotFoundError:
        raise MlflowException(
            f"Failed to find answer similarity metric for version {metric_version}."
            f" Please check the version",
            error_code=INVALID_PARAMETER_VALUE,
        ) from None
    except Exception as e:
        raise MlflowException(
            f"Failed to construct answer similarity metric {metric_version}. Error: {e!r}",
            error_code=INTERNAL_ERROR,
        ) from None

    if examples is None:
        examples = answer_similarity_class_module.default_examples
    if model is None:
        model = answer_similarity_class_module.default_model

    return make_genai_metric(
        name="answer_similarity",
        definition=answer_similarity_class_module.definition,

View on GitHub (pinned to 6a27f2decc)

Solutions

Run mlflow.__version__ and use a metric_version that exists in mlflow/metrics/genai/prompts/ of that install
Omit metric_version to use _get_latest_metric_version()'s default
Upgrade mlflow if you need a newer prompt version
Fix case/typo in the version string (versions are lowercase like 'v1')

Example fix

// before
answer_similarity(metric_version="V2", model="openai:/gpt-4o-mini")
// after
answer_similarity(metric_version="v1", model="openai:/gpt-4o-mini")

Defensive patterns

Strategy: validation

Validate before calling

import mlflow, importlib.util
if metric_version and not importlib.util.find_spec(f"mlflow.metrics.genai.prompts.{metric_version}"):
    raise ValueError(f"metric_version {metric_version} not in mlflow {mlflow.__version__}")

Type guard

def known_metric_version(v):
    return v is None or (isinstance(v, str) and importlib.util.find_spec(f"mlflow.metrics.genai.prompts.{v}") is not None)

Try / catch

from mlflow.exceptions import MlflowException
try:
    m = answer_similarity(metric_version=v, model=model)
except MlflowException as e:
    if e.error_code == "INVALID_PARAMETER_VALUE":
        m = answer_similarity(model=model)  # fall back to latest default
    else:
        raise

Prevention

Omit metric_version unless you specifically need a pinned prompt version
Verify the version exists in your installed mlflow before referencing it in code
Pin mlflow in requirements to keep versions and code in sync

When it happens

Trigger: Calling answer_similarity(metric_version='vX') where prompts/vX does not exist — a typo like 'V1' vs 'v1', a version from docs of a newer MLflow, or metric_version=None defaulting to a version not present in the installed mlflow.

Common situations: Following a blog/tutorial referencing a prompt version not in your installed mlflow; downgrading mlflow while keeping code that names a newer version; case typos in the version string.

Related errors

AI-assisted analysis of mlflow/mlflow@6a27f2decc (2026-08-29). Data as JSON: /api/errors/254a921146ccbc95. Report an issue: GitHub.