MlflowApi (default)

pydantic model mlopus.mlflow.providers.mlflow.MlflowApi

Bases: BaseMlflowApi

MLflow API provider based on open source MLflow.

Plugin name: mlflow

Requires extras: mlflow

Default cache dir: ~/.cache/mlopus/mlflow-providers/mlflow/<hashed-tracking-uri>

Assumptions:

• No artifacts proxy.

• SQL database is server-managed.

field tracking_uri: str = None

MLflow server URL or path to a local directory. Defaults to the environment variable MLFLOW_TRACKING_URI, falls back to ~/.cache/mlflow.

field healthcheck: bool = True

If true and not in offline_mode, eagerly attempt connection to the server after initialization.

field client_settings: Dict[str, str | int] [Optional]

MLflow client settings. Keys are like the open-source MLflow environment variables, but lower case and without the MLFLOW_ prefix. Example: http_request_max_retries. See: https://mlflow.org/docs/latest/python_api/mlflow.environment_variables.html

field tag_keys: MlflowTagKeys [Optional]

Tag keys for storing internal information such as parent run ID.

field query_push_down: MlflowQueryPushDown [Optional]

Utility for partial translation of MongoDB queries to open-source MLflow SQL. Users may replace this with a different implementation when subclassing the API.

field data_translation: MlflowDataTranslation [Optional]

Utility for translating keys and values from MLOpus schema to native MLflow schema and back. Users may replace this with a different implementation when subclassing the API.