tfma.default_eval_shared_model

Returns default EvalSharedModel.

tfma.default_eval_shared_model(
    eval_saved_model_path: str,
    add_metrics_callbacks: Optional[List[types.AddMetricsCallbackType]] = None,
    include_default_metrics: Optional[bool] = True,
    example_weight_key: Optional[Union[str, Dict[str, str]]] = None,
    additional_fetches: Optional[List[str]] = None,
    blacklist_feature_fetches: Optional[List[str]] = None,
    tags: Optional[List[str]] = None,
    model_name: str = '',
    eval_config: Optional[tfma.EvalConfig] = None,
    custom_model_loader: Optional[tfma.types.ModelLoader] = None,
    rubber_stamp: Optional[bool] = False
) -> tfma.types.EvalSharedModel

Used in the notebooks

Used in the guide Used in the tutorials

eval_saved_model_path Path to EvalSavedModel.
add_metrics_callbacks Optional list of callbacks for adding additional metrics to the graph (see EvalSharedModel for more information on how to configure additional metrics). Metrics for example count and example weights will be added automatically. Only used if EvalSavedModel used.
include_default_metrics DEPRECATED. Use eval_config.options.include_default_metrics.
example_weight_key DEPRECATED. Use eval_config.model_specs.example_weight_key or eval_config.model_specs.example_weight_keys.
additional_fetches Optional prefixes of additional tensors stored in signature_def.inputs that should be fetched at prediction time. The "features" and "labels" tensors are handled automatically and should not be included. Only used if EvalSavedModel used.
blacklist_feature_fetches Optional list of tensor names in the features dictionary which should be excluded from the fetches request. This is useful in scenarios where features are large (e.g. images) and can lead to excessive memory use if stored. Only used if EvalSavedModel used.
tags Optional model tags (e.g. 'serve' for serving or 'eval' for EvalSavedModel).
model_name Optional name of the model being created (should match ModelSpecs.name). The name should only be provided if multiple models are being evaluated.
eval_config Eval config.
custom_model_loader Optional custom model loader for non-TF models.
rubber_stamp True when this run is a first run without a baseline model while a baseline is configured, the diff thresholds will be ignored.