mlflow/mlflow · error · MlflowException

INVALID_PARAMETER_VALUE

INVALID_PARAMETER_VALUE

Error message

Searching traces by model_id is not supported on the current tracking server.

What it means

RestStore.search_traces refuses any call that passes model_id, because filtering traces by model association is not implemented for remote (REST) tracking servers. It raises INVALID_PARAMETER_VALUE immediately instead of sending an unsupported filter to the server.

Source

Thrown at mlflow/store/tracking/rest_store.py:591

            req_body,
            endpoint=f"{_V3_TRACE_REST_API_PATH_PREFIX}/batchGetInfos",
        )
        return [TraceInfo.from_proto(proto) for proto in response_proto.trace_infos]

    def search_traces(
        self,
        experiment_ids: list[str] | None = None,
        filter_string: str | None = None,
        max_results: int = SEARCH_TRACES_DEFAULT_MAX_RESULTS,
        order_by: list[str] | None = None,
        page_token: str | None = None,
        model_id: str | None = None,
        locations: list[str] | None = None,
    ):
        locations = _resolve_experiment_ids_and_locations(experiment_ids, locations)

        if model_id is not None:
            raise MlflowException.invalid_parameter_value(
                "Searching traces by model_id is not supported on the current tracking server.",
            )

        return self._search_traces(
            locations=locations,
            filter_string=filter_string,
            max_results=max_results,
            order_by=order_by,
            page_token=page_token,
        )

    def _search_traces(
        self,
        locations: list[str],
        filter_string: str | None = None,
        max_results: int = SEARCH_TRACES_DEFAULT_MAX_RESULTS,
        order_by: list[str] | None = None,
        page_token: str | None = None,

View on GitHub (pinned to 6a27f2decc)

Solutions

1. Remove the model_id argument and filter traces with a filter_string on trace attributes/tags instead.
2. Use the logged-model trace search API if your MLflow version supports it server-side.
3. Use a local SQLAlchemy tracking store where model_id filtering is implemented, or upgrade the server to a version that supports it.

Example fix

// before
client.search_traces(experiment_ids=["1"], model_id="m-123")
// after
client.search_traces(experiment_ids=["1"], filter_string="tags.model_id = 'm-123'")

def assert_no_model_id(**kwargs):
    if kwargs.get("model_id") is not None:
        raise ValueError("REST store search_traces does not support model_id")

try:
    client.search_traces(experiment_ids=eids, model_id=mid)
except MlflowException as e:
    if e.error_code == "INVALID_PARAMETER_VALUE":
        pass  # retry without model_id or with a filter_string

When it happens

Trigger: Calling MlflowClient.search_traces(..., model_id="<id>") against a REST/Databricks tracking store — i.e., any non-local backend.

Common situations: Code that works against a local SQLAlchemy store (which supports model_id filtering) fails when pointed at a remote server; notebooks mixing logged-model queries with tracing on Databricks.

Related errors

• Invalid experiment ID format: {exp_id}. Error: {e!s}
• log_spans is not supported: could not identify MLflow server
• log_spans is not supported: MLflow server version {server_ve
• Artifact location not found in trace tags
• Invalid response format: missing trace_info: ${JSON.stringif

AI-assisted analysis of mlflow/mlflow@6a27f2decc (2026-08-29). Data as JSON: /api/errors/0e104b098bd6f029. Report an issue: GitHub.